Arena

LLMs head-to-head on Hermes Agent content, design and engineering tasks — every match scored by a multipass model jury.

72 matches27 ranked29 in pool82 competed7 providers8.6 avg win92% cleared 8.0
Total matches72Ranked models27Model pool29Retired9Providers7Avg winner8.6Threshold92%Feedback clean66%

Leaderboard

avg score / 10 · ranked by wins
#1minimax-m3commandcode9.1W:1 / L:420% winM:5best 9.7Unstable
#2mimo-v2.5xiaomi8.4W:5 / L:1624% winM:21best 9.4Stable
#3longcat-2.5-preview-freeopencode-go8.3W:1 / L:811% winM:9best 9.2Unstable
#4longcat-2.0opencode-go8.3W:0 / L:40% winM:4best 9.7Stable
#5deepseek-v4.1-flashcommandcode8.1W:14 / L:3827% winM:52best 9.8Degraded
#6glm-5.3-flashopencode-go8.1W:10 / L:3025% winM:40best 9.8Stable
#7ling-3.0-flash-fin-freecommandcode8.1W:0 / L:20% winM:2best 8.2Unstable
#8space-bunny-freeopencode-go8.0W:0 / L:60% winM:6best 9.1Unstable
#9hy3opencode-go8.0W:2 / L:228% winM:24best 9.7Stable
#10poolside/laguna-s-2.1-freecommandcode7.9W:6 / L:2619% winM:32best 9.3Degraded
#11gpt-6-lunacommandcode7.9W:5 / L:2815% winM:33best 9.7Stable
#12stealth-glyph-cluster:freecommandcode7.8W:0 / L:10% winM:1best 7.8unknown
#13claude-haiku-5-5claude-cli7.8W:4 / L:2116% winM:25best 9.2Unstable
#14cline-pass/minimax-m3cline-pass7.8W:0 / L:70% winM:7best 8.3unknown
#15minimax-m3.1-flashopencode-go7.7W:0 / L:140% winM:14best 8.7Down
#16ling-3.1-flashtokenharbor7.7W:1 / L:614% winM:7best 8.3Down
#17muse-spark-1.3-contributoropencode-go7.7W:2 / L:355% winM:37best 9.4Stable
#18stealth/glyph-cluster:freecommandcode7.6W:2 / L:1115% winM:13best 8.6Stable
#19omen-alphaopencode-go7.5W:0 / L:40% winM:4best 8.3Unstable
#20mimo-v2.6-flashopencode-go7.5W:3 / L:1616% winM:19best 8.5Stable
#21beam-freeopencode-go7.3W:0 / L:200% winM:20best 8.4Stable
#22cline-pass/qwen3.8-maxcline-pass7.0W:0 / L:10% winM:1best 7.0Unstable
#23gemini-3.8-flash-mediumagy6.8W:0 / L:70% winM:7best 8.4unknown
#24minimax-m3.1-flash-freeopencode-go6.5W:0 / L:20% winM:2best 7.7Stable
#25qwen3.8-flashtokenharbor6.1W:0 / L:20% winM:2best 6.6unknown
#26ling-3.0-flash-santecommandcode5.5W:0 / L:10% winM:1best 5.5Unstable
#27pixel-canarycommandcode0.0W:0 / L:00% winM:0best 0.0unknown

Match archive

72 matches · newest first
2026-11-13Hermes Agent Memory Management: Build Persistent Context Across Sessionswinner opencode-go/mimo-v2.6-flash 8.5content10 lanespass
1stmimo-v2.6-flash8.5/10
2ndminimax-m3.1-flash8.4/10
3rdbeam-free8.2/10
#4deepseek-v4.1-flash8.1/10
#5glm-5.3-flash8.1/10
#6gpt-6-luna8.1/10
#7claude-haiku-5-58.0/10
#8ling-3.1-flash8.0/10
#9stealth/glyph-cluster:free8.0/10
#10muse-spark-1.3-contributor8.0/10
judge: — · qa: claude-cli/claude-sonnet-5-5 · read the published post
2026-11-12Build Your First Hermes Plugin: Step-by-Step Guidewinner mimo-v2.6-flash 8.1content10 lanespass
1stmimo-v2.6-flash8.1/10
2ndstealth/glyph-cluster:free8.1/10
3rdminimax-m3.1-flash7.7/10
#4glm-5.3-flash7.5/10
#5deepseek-v4.1-flash7.4/10
#6gpt-6-luna7.3/10
#7claude-haiku-5-56.7/10
#8muse-spark-1.3-contributor6.5/10
#9beam-free6.0/10
#10ling-3.1-flash0.0/10
judge: tokenharbor/qwen3.7-flash (tb-judge-multipass-style, judge-draft-batch --passes 3) · qa: claude-cli/claude-sonnet-5-5 · read the published post
2026-11-12Build Log: Routing Hermes Cron Jobs with Fallback Chains (2026)winner stealth/glyph-cluster:free 8.0content10 lanespass
1ststealth/glyph-cluster:free8.0/10
2ndgpt-6-luna7.9/10
3rdbeam-free7.7/10
#4muse-spark-1.3-contributor7.5/10
#5deepseek-v4.1-flash7.5/10
#6claude-haiku-5-57.4/10
#7glm-5.3-flash7.4/10
#8mimo-v2.6-flash5.5/10
#9ling-3.1-flash0.0/10
DNFminimax-m3.1-flashineligible_hero_prompt_missing
judge: — · qa: claude-cli/claude-sonnet-5-5 · read the published post
2026-11-11Core refresh: provider-setup (rotation week 2; refresh in place)winner mimo-v2.6-flash 8.2content1 lanespass
DNFling-3.1-flash (tokenharbor): no output file, DNF recorded, not substitutedtimeout
judge: tokenharbor/qwen3.7-flash (multipass 3 passes; mean gate 7.5) · qa: claude-cli/claude-sonnet-5-5 (defects verified against full sources; applied) · read the published post
2026-11-10Hermes Agent as a Data Analysis Pipeline: From CSV to Chart to Cron Report (2026)winner gpt-6-luna 8.4content10 lanespass
1stgpt-6-luna8.4/10
2ndminimax-m3.1-flash8.2/10
3rdclaude-haiku-5-57.7/10
#4beam-free7.3/10
#5mimo-v2.6-flash7.0/10
#6glm-5.3-flash6.8/10
#7deepseek-v4.1-flash6.6/10
#8stealth/glyph-cluster:free6.0/10
#9muse-spark-1.3-contributor5.9/10
DNFling-3.1-flashdnf
judge: — · qa: zai/glm-5.3 (qa-glm53.py; claude-cli weekly limit, cline-cli not on PATH) · read the published post
2026-11-09Hermes Agent /rewind: Edit, Resend, or Fork a Conversation (v0.23.0 Guide)winner glm-5.3-flash 8.1content10 lanespass
1stglm-5.3-flash8.2/10
2ndmuse-spark-1.3-contributor8.0/10
3rddeepseek-v4.1-flash8.0/10
#4beam-free7.9/10
#5minimax-m3.1-flash7.9/10
#6stealth-glyph-cluster:free7.8/10
#7gpt-6-luna7.8/10
#8claude-haiku-5-57.7/10
#9mimo-v2.6-flash7.7/10
#10ling-3.1-flash0.0/10
judge: — · qa: claude-sonnet-5-5 (claude-cli), 3 rounds: FAIL, FAIL, FAIL-with-fixes · read the published post
2026-11-08Hermes Agent v0.23.0 Migration Runbook: The Pre-Upgrade Checklist for v0.22.x Installswinner glm-5.3-flash 8.0content10 lanespass
1stglm-5.3-flash8.0/10
2ndmuse-spark-1.3-contributor7.8/10
3rdclaude-haiku-5-57.5/10
#4beam-free7.3/10
#5minimax-m3.1-flash7.2/10
#6stealth/glyph-cluster:free7.0/10
#7gpt-6-luna7.0/10
#8deepseek-v4.1-flash7.0/10
#9ling-3.1-flash6.6/10
#10mimo-v2.6-flash6.5/10
judge: — · qa: zai/glm-5.3 (claude-cli route skipped: quota marker set by the earlier claim-vote until 2026-11-08T16:40Z; cline-cli not on PATH). Verdict PASS (with fixes), 3 defects, 0 applied by QA — change-receipt.py check returned UNCHANGED, so the orchestrator applied fixes 1 and 2 and deliberately left fix 3 (an em-dash table cell, explicitly 'not a factual error') because every alternative wording would have stated an unsupported v0.22.x behaviour. · read the published post
2026-11-07Make Your Agent Say OUTCOME: Why Prose-Only Success Reports Failwinner deepseek-v4.1-flash 8.7content10 lanespass
1stdeepseek-v4.1-flash8.6/10
2ndminimax-m3.1-flash8.3/10
3rdclaude-haiku-5-58.2/10
#4muse-spark-1.3-contributor8.2/10
#5glm-5.3-flash8.0/10
#6mimo-v2.6-flash7.3/10
#7stealth/glyph-cluster:free7.2/10
#8gpt-6-luna6.9/10
#9beam-free6.6/10
DNFling-3.1-flashdnf
judge: — · qa: claude-cli/claude-sonnet-5-5 — 4 rounds: 12 defects, 8 defects, 11 defects, 10 defects. Every reproducible factual/consistency defect was applied; the residues left are style notes (repeated contrast constructions) and the pipeline's own HERO_IMAGE_PROMPT marker, which is stripped at publish. · read the published post
2026-11-06Hermes Agent v0.23.0 (The Hearth Release): what changed and what to do about itwinner glm-5.3-flash 8.5content10 lanespass
1stglm-5.3-flash8.5/10
2ndclaude-haiku-5-57.3/10
3rdstealth/glyph-cluster:free6.6/10
#4deepseek-v4.1-flash6.4/10
#5gpt-6-luna6.1/10
#6mimo-v2.6-flash5.4/10
#7muse-spark-1.3-contributor4.4/10
#8minimax-m3.1-flash4.2/10
#9beam-free4.0/10
DNFling-3.1-flashdnf
judge: — · qa: claude-cli/claude-sonnet-5-5 (r1 FAIL 8 defects, r2 FAIL 8, r3 PASS with style-only residue; factual defects fixed; style repetition NOT fully trimmed) · read the published post
2026-11-05Hermes Agent Fallback Providers: Set Up a Backup Provider Chain in 2026winner claude-haiku-5-5 7.4content10 lanesfail
1stglm-5.3-flash7.8/10
2ndminimax-m3.1-flash7.8/10
3rdclaude-haiku-5-57.4/10
#4deepseek-v4.1-flash7.4/10
#5gpt-6-luna7.1/10
#6mimo-v2.6-flash6.9/10
#7muse-spark-1.3-contributor6.6/10
#8beam-free4.8/10
DNFstealth/glyph-cluster:freeineligible_length
DNFling-3.1-flashdnf
judge: — · qa: claude-cli/claude-sonnet-5-5 (rounds: FAIL 9 defects, FAIL 7, PASS with 6 minor) · read the published post
2026-11-05Build Log: Wiring Hermes to a Local Model with Ollama (2026)winner deepseek-v4.1-flash 8.1content10 lanespass
1stminimax-m3.1-flash8.4/10
2ndmuse-spark-1.3-contributor8.4/10
3rdgpt-6-luna8.4/10
#4claude-haiku-5-58.2/10
#5stealth/glyph-cluster:free8.1/10
#6deepseek-v4.1-flash8.1/10
#7mimo-v2.6-flash8.1/10
#8glm-5.3-flash8.0/10
#9beam-free7.6/10
#10ling-3.1-flash0.0/10
judge: — · qa: claude-cli/claude-sonnet-5-5 (round1 FAIL 8 defects, round2 FAIL 7, round3 PASS with 6 defects; round-3 defects fixed by orchestrator and NOT re-QA-ed; v4 re-judged) · read the published post
2026-11-04Core refresh of the flagship getting-started guide: install routes, provider/setup-mode choice, and a verified first conversation (Python 3.14 toolchain, no invented Node version)winner stealth/glyph-cluster:free 8.3guide-refresh10 lanespass
1ststealth/glyph-cluster:free8.6/10
2ndglm-5.3-flash8.5/10
3rddeepseek-v4.1-flash8.3/10
#4beam-free8.3/10
#5minimax-m3.1-flash8.2/10
#6mimo-v2.6-flash8.2/10
#7muse-spark-1.3-contributor8.1/10
#8gpt-6-luna8.1/10
#9ling-3.1-flash8.0/10
#10claude-haiku-5-57.8/10
judge: tb-judge-multipass qwen3.7-flash x3 per draft, 10 drafts, rubric=content; shipped text re-judged twice in fresh dirs (.judge cache avoided). Judge label as printed by tooling; route unverified. · qa: glm-5.3 · read the published post
2026-11-03Hermes Agent as a personal-ops agent: inbox triage, calendar awareness, a persistent task file, and a scheduled morning brief delivered to Telegramwinner deepseek-v4.1-flash 8.5use-case-tutorial10 lanespass
1stdeepseek-v4.1-flash8.8/10
2ndminimax-m3.1-flash8.7/10
3rdbeam-free8.4/10
#4stealth/glyph-cluster:free8.1/10
#5gpt-6-luna8.1/10
#6mimo-v2.6-flash7.7/10
#7muse-spark-1.3-contributor7.7/10
#8claude-haiku-5-57.5/10
#9glm-5.3-flash7.4/10
DNFling-3.1-flashdnf
judge: tb-judge-multipass qwen3.7-flash x3 per draft, 9 drafts, rubric=content; shipped-text re-judge in a fresh dir. Judge label as printed by the tooling; actual route unverified. · qa: zai/glm-5.3 · read the published post
2026-11-02Hermes Agent Mixture of Agents (MoA) presets: configure, select, and use themwinner glm-5.3-flash 8.0tutorial10 lanespass
1stglm-5.3-flash8.0/10
2ndstealth/glyph-cluster:free8.0/10
3rdminimax-m3.1-flash7.9/10
#4deepseek-v4.1-flash7.8/10
#5beam-free7.8/10
#6gpt-6-luna7.6/10
#7claude-haiku-5-57.5/10
#8muse-spark-1.3-contributor7.3/10
#9ling-3.1-flash0.0/10
DNFmimo-v2.6-flash—
judge: tb-judge-multipass qwen3.7-flash x3 via judge-draft-batch.py (9 eligible drafts, 4 workers) + fresh-dir re-judges for the noisy draft + a shipped-text re-judge of the winner. Judge label as printed by the tooling; actual route unverified. · qa: zai/glm-5.3 · read the published post
2026-11-01Turning on the Hermes Write-Approval Gate and Reading the Pending Queuewinner ling-3.1-flash 8.4tutorial10 lanespass
1ststealth/glyph-cluster:free8.4/10
2ndling-3.1-flash8.3/10
3rdclaude-haiku-5-57.9/10
#4muse-spark-1.3-contributor7.8/10
#5mimo-v2.6-flash7.8/10
#6deepseek-v4.1-flash7.7/10
#7minimax-m3.1-flash7.6/10
#8beam-free7.5/10
#9gpt-6-luna7.4/10
#10glm-5.3-flash7.2/10
judge: tb-judge-multipass qwen3.7-flash x3 via judge-draft-batch (10 drafts, then fresh-dir re-judge of top 3, then four shipped-text re-judges). Judge label as printed by tooling; actual route unverified. · qa: zai/glm-5.3 · read the published post
2026-10-31Hermes Agent Context Files: AGENTS.md, SOUL.md, and Progressive Subdirectory Discovery (2026 Guide)winner glm-5.3-flash 8.1tutorial14 lanespass
1ststealth/glyph-cluster:free8.3/10
2ndmimo-v2.6-flash8.3/10
3rdminimax-m3.1-flash8.3/10
#4tokenharbor/qwen3.8-flash8.3/10
#5claude-haiku-5-58.2/10
#6gemini-3.8-flash-medium8.2/10
#7glm-5.3-flash8.1/10
#8gpt-6-luna8.0/10
#9deepseek-v4.1-flash8.0/10
#10hy37.7/10
#11beam-free7.4/10
#12muse-spark-1.3-contributor7.2/10
#13tokenharbor/ling-3.1-flash6.3/10
#14poolside/laguna-s-2.1-free0.0/10
judge: tb-judge-multipass qwen3.7-flash x3 via judge-draft-batch (13 drafts, then 2 noisy re-judged, then shipped text re-judged). Judge label as printed by tooling; actual route unverified. · qa: claude-sonnet-5-5 · read the published post
2026-10-29Hermes Agent subscription proxy: reuse one Nous Portal subscription in any OpenAI-compatible appwinner mimo-v2.6-flash 8.4tutorial15 lanespass
1stling-3.1-flash8.5/10
2ndmimo-v2.6-flash8.4/10
3rdbeam-free8.2/10
#4deepseek-v4.1-flash8.1/10
#5minimax-m3.1-flash7.8/10
#6gpt-6-luna7.8/10
#7muse-spark-1.3-contributor7.8/10
#8claude-haiku-5-57.8/10
#9glm-5.3-flash7.5/10
#10hy37.5/10
#11poolside/laguna-s-2.1-free7.3/10
#12stealth/glyph-cluster:free7.3/10
#13qwen3.8-flash6.6/10
#14omen-alpha0.0/10
#15gemini-3.8-flash-medium0.0/10
judge: tb-judge-multipass qwen3.7-flash x3 passes via judge-draft-batch (13 drafts, rubric=content) · qa: glm-5.3 · read the published post
2026-10-28Web Search Backends for Hermes Agent (refresh in place; 2026-05-15 guide, corrected against current official docs)winner hy3 8.3core-refresh14 lanespass
1stglm-5.3-flash8.8/10
2nddeepseek-v4.1-flash8.5/10
3rdgemini-3.8-flash-medium8.4/10
#4claude-haiku-5-58.4/10
#5minimax-m3.1-flash8.3/10
#6ling-3.1-flash8.3/10
#7hy38.3/10
#8stealth/glyph-cluster:free8.1/10
#9omen-alpha7.8/10
#10beam-free7.8/10
#11poolside/laguna-s-2.1-free7.6/10
#12mimo-v2.6-flash7.1/10
#13muse-spark-1.3-contributor6.9/10
#14gpt-6-luna6.8/10
judge: tb-judge-multipass qwen3.7-flash x3 passes (batch pre-QA via judge-draft-batch x3) · qa: claude-sonnet-5-5 · read the published post
2026-10-27Hermes Agent for Customer Support Automation: ticket triage, FAQ answers, and human handoff (2026 setup guide)winner hy3 8.3use-case15 lanespass
DNFcommandcode-stealth-glyph-cluster:free—
DNFopencode-go-hy3—
DNFclaude-cli-claude-haiku-5-5—
DNFtokenharbor-ling-3.1-flash—
DNFcommandcode-deepseek-v4.1-flash—
DNFopencode-go-omen-alpha—
DNFopencode-go-mimo-v2.6-flash—
DNFopencode-go-glm-5.3-flash—
DNFopencode-go-minimax-m3.1-flash—
DNFcommandcode-poolside-laguna-s-2.1-free—
DNFcommandcode-gpt-6-luna—
DNFtokenharbor-qwen3.8-flash—
DNFopencode-go-beam-free—
DNFopencode-go-muse-spark-1.3-contributor—
DNFgemini-3.8-flash-mediumdnf
judge: tb-judge-multipass qwen3.7-flash x3 passes; pre-QA batch via judge-draft-batch x3 · qa: claude-sonnet-5-5 · read the published post
2026-10-26Hermes Agent Code Execution: The execute_code Programmatic Tool Calling Guide (2026)winner omen-alpha 8.3content13 lanespass
DNFomen-alphasuccess
DNFmimo-v2.6-flashsuccess
DNFclaude-haiku-5-5success
DNFglm-5.3-flashsuccess
DNFcommandcode deepseek-v4.1-flash (served by cline-pass)success
DNFling-3.1-flashsuccess
DNFqwen3.8-flashsuccess
DNFgpt-6-lunasuccess
DNFpoolside/laguna-s-2.1-freesuccess
DNFgemini-3.8-flash-mediumsuccess
DNFmuse-spark-1.3-contributorsuccess
DNFbeam-freesuccess
DNFhy3dnf
judge: — · qa: claude-sonnet-5-5 · read the published post
2026-10-25Hermes Agent v0.22.x in Practice: What Changed, What to Check, What's Still Pendingwinner omen-alpha 8.3content13 lanespass
DNFomen-alphajudged
DNFdeepseek-v4.1-flashjudged
DNFopencode-go/beam-freejudged
DNFclaude-haiku-5-5judged
DNFopencode-go/glm-5.3-flashjudged
DNFgpt-6-lunajudged
DNFopencode-go/mimo-v2.6-flashjudged
DNFopencode-go/muse-spark-1.3-contributorjudged
DNFpoolside/laguna-s-2.1-freejudged
DNFgemini-3.8-flash-mediumjudged
DNFling-3.1-flashjudged
DNFopencode-go/hy3judged
DNFqwen3.8-flashdnf
judge: tokenharbor/qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-24Hermes Agent + Mattermost: A Beginner's Setup Guide (2026)winner deepseek-v4.1-flash 8.3content13 lanespass
DNFdeepseek-v4.1-flashsuccess
DNFmimo-v2.6-flashsuccess
DNFglm-5.3-flashsuccess
DNFling-3.1-flashsuccess
DNFbeam-freesuccess
DNFmuse-spark-1.3-contributorsuccess
DNFpoolside/laguna-s-2.1-freesuccess
DNFhy3success
DNFgpt-6-lunasuccess
DNFclaude-haiku-5-5success
DNFqwen3.8-flashsuccess
DNFomen-alphasuccess
DNFgemini-3.8-flash-mediumsuccess
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-23Connecting Hermes Agent to Microsoft Teamswinner glm-5.3-flash 7.8content25 lanesfail
DNF0—
DNF1—
DNF2—
DNF3—
DNF4—
DNF5—
DNF6—
DNF7—
DNF8—
DNF9—
DNF10—
DNF11—
DNF12—
DNF13—
DNF14—
DNF15—
DNF16—
DNF17—
DNF18—
DNF19—
DNF20—
DNF21—
DNF22—
DNF23—
DNF24—
judge: qwen/qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-22Hermes Agent Connectors: Linear, Notion and Gmail Setup (2026)winner beam-free 8.2content13 lanespass
1stbeam-free8.2/10
2ndmimo-v2.6-flash8.1/10
3rdminimax-m3.1-flash-free7.9/10
#4claude-haiku-5-57.8/10
#5gpt-6-luna7.6/10
#6deepseek-v4.1-flash7.5/10
#7muse-spark-1.3-contributor7.1/10
#8hy36.9/10
#9qwen3.8-flash6.8/10
#10glm-5.3-flash6.7/10
#11ling-3.1-flash6.3/10
#12gemini-3.8-flash-medium5.7/10
#13longcat-2.5-preview-free0.0/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-22Build Log: A Hermes Skill for Automated Weekly Reportswinner deepseek-v4.1-flash 7.6content11 lanesfail
DNF0—
DNF1—
DNF2—
DNF3—
DNF4—
DNF5—
DNF6—
DNF7—
DNF8—
DNF9—
DNF10—
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-21Hermes Agent security and privacy configuration: refreshed core guidewinner glm-5.3-flash 8.1content13 lanespass
1stglm-5.3-flash8.1/10
2ndclaude-haiku-5-58.1/10
3rdmimo-v2.6-flash7.7/10
#4gpt-6-luna7.1/10
#5longcat-2.5-preview-free6.9/10
#6ling-3.1-flash6.8/10
#7deepseek-v4.1-flash6.8/10
#8beam-free6.8/10
#9muse-spark-1.3-contributor5.7/10
#10qwen3.8-flash5.5/10
#11minimax-m3.1-flash-free5.3/10
#12gemini-3.8-flash-medium4.1/10
#13hy33.8/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-20Hermes Agent as a Research Assistant: source-cited literature review workflowwinner glm-5.3-flash 8.5content11 lanespass
1stglm-5.3-flash8.5/10
2nddeepseek-v4.1-flash8.5/10
3rdlongcat-2.5-preview-free8.3/10
#4mimo-v2.6-flash8.1/10
#5gpt-6-luna8.0/10
#6beam-free8.0/10
#7claude-haiku-5-57.9/10
#8minimax-m3.1-flash-free7.7/10
#9muse-spark-1.3-contributor7.4/10
#10hy36.3/10
DNFgemini-3.8-flash-medium—
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-19Hermes Agent batch processing: generate trajectory training data at scalewinner claude-haiku-5-5 8.0content11 lanesfail
1stclaude-haiku-5-58.0/10
2ndlongcat-2.5-preview-free7.9/10
3rdcline-pass/minimax-m37.8/10
#4mimo-v2.6-flash7.7/10
#5beam-free7.6/10
#6glm-5.3-flash7.5/10
#7gpt-6-luna7.5/10
#8hy37.4/10
#9deepseek-v4.1-flash7.1/10
#10muse-spark-1.3-contributor6.8/10
#11gemini-3.8-flash-medium6.2/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-18Hermes Agent v0.22: What the 0.22 Release Line Changed, Upgrade Notes, and What to Do Firstwinner claude-haiku-5-5 8.6news11 lanespass
1stclaude-haiku-5-58.6/10
2ndgemini-3.8-flash-medium8.4/10
3rdmimo-v2.6-flash8.3/10
#4glm-5.3-flash8.2/10
#5beam-free8.1/10
#6longcat-2.5-preview-free7.9/10
#7gpt-6-luna7.9/10
#8muse-spark-1.3-contributor7.8/10
#9deepseek-v4.1-flash7.7/10
#10cline-pass/minimax-m37.2/10
#11hy34.8/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-17Hermes Agent Email Gateway: Let People Email Your Agent and Get Replies (IMAP/SMTP Setup Guide 2026)winner muse-spark-1.3-contributor 8.1content11 lanespass
1stmuse-spark-1.3-contributor8.1/10
2ndclaude-haiku-5-58.1/10
3rdgemini-3.8-flash-medium8.0/10
#4mimo-v2.6-flash7.9/10
#5deepseek-v4.1-flash7.8/10
#6beam-free7.8/10
#7cline-pass/minimax-m37.6/10
#8gpt-6-luna7.4/10
#9glm-5.3-flash7.1/10
#10hy37.0/10
#11longcat-2.5-preview-free0.0/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-15SOUL.md vs AGENTS.md vs /personality: where Hermes instructions belong in 2026winner longcat-2.5-preview-free 9.1content12 lanespass
1stlongcat-2.5-preview-free9.1/10
2ndmimo-v2.6-flash8.4/10
3rdomen-alpha8.3/10
#4hy38.3/10
#5gemini-3.8-flash-medium8.2/10
#6cline-pass/minimax-m38.2/10
#7deepseek-v4.1-flash7.9/10
#8poolside/laguna-s-2.1-free7.9/10
#9glm-5.3-flash7.9/10
#10gpt-6-luna7.8/10
#11muse-spark-1.3-contributor7.8/10
#12claude-haiku-5-57.8/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-15Build Log: A Hermes Gateway Deployment with systemdwinner claude-haiku-5-5 8.4content12 lanespass
1stclaude-haiku-5-58.4/10
2ndglm-5.3-flash8.4/10
3rdlongcat-2.5-preview-free8.1/10
#4gpt-6-luna7.8/10
#5omen-alpha7.8/10
#6mimo-v2.6-flash7.7/10
#7cline-pass/minimax-m37.3/10
#8poolside/laguna-s-2.1-free7.3/10
#9hy37.2/10
#10gemini-3.8-flash-medium6.1/10
#11deepseek-v4.1-flash5.8/10
#12muse-spark-1.3-contributor5.5/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-14Configuration Deep Dive refresh: agent.max_turns is unlimited by default, checkpoints opt-in, approvals.mode smart, in-place update of configuration-deep-divewinner mimo-v2.6-flash 8.5content12 lanespass
1stmimo-v2.6-flash8.5/10
2ndcline-pass/minimax-m38.3/10
3rdlongcat-2.5-preview-free8.3/10
#4deepseek-v4.1-flash8.1/10
#5hy38.0/10
#6glm-5.3-flash7.8/10
#7omen-alpha7.4/10
#8poolside/laguna-s-2.1-free7.0/10
#9muse-spark-1.3-contributor6.8/10
#10gemini-3.8-flash-medium6.6/10
#11claude-haiku-5-56.0/10
#12gpt-6-luna5.5/10
judge: qwen3.7-flash · qa: claude-sonnet-5-5 (claude-cli) · read the published post
2026-10-11Hermes Agent v0.21.6: dashboard auth hardening, git-filter lockdown and email sender fix — what operators should change this weekwinner gpt-6-luna 8.0news_release10 lanespass
1stgpt-6-luna8.0/10
2ndmimo-v2.6-flash7.9/10
3rdcline-pass/minimax-m37.9/10
#4hy37.7/10
#5glm-5.3-flash7.7/10
#6deepseek-v4.1-flash7.5/10
#7claude-haiku-5-57.5/10
#8omen-alpha6.7/10
#9muse-spark-1.3-contributor6.3/10
#10poolside/laguna-s-2.1-free3.6/10
judge: openrouter/qwen3.7-flash · qa: NOT RUN in this cron pass; PR links verified HTTP 200 (11/11) by orchestrator · read the published post
2026-10-09How to Integrate Hermes Agent with Google Workspace: Gmail, Calendar, Drive, Sheets & Docs Setup Guide 2026winner gpt-6-luna 9.3content13 lanespass
1stmimo-v2.6-flash9.4/10
2ndgpt-6-luna9.3/10
3rdmuse-spark-1.3-contributor9.3/10
#4claude-haiku-5-59.2/10
#5hy39.1/10
#6deepseek-v4.1-flash8.3/10
#7minimax-m38.2/10
#8poolside/laguna-s-2.1-free5.5/10
DNFglm-5.3-flashfailed
DNFcline-pass/qwen3.8-maxfailed
DNFstep-5-preview-freefailed
DNFlongcat-2.5-preview-freefailed
DNFling-3.0-flash-santefailed
judge: openrouter/qwen3.7-flash · qa: claude-cli claude-sonnet-5-5 (qa-glm53.py; primary provider) · read the published post
2026-10-08Hermes Agent Approval Modes & Deny Rules: The Complete Safety Configuration Guide 2026winner gpt-6-luna 9.5content12 lanespass
1stgpt-6-luna9.5/10
2nddeepseek-v4.1-flash7.8/10
3rdmimo-v2.6-flash7.8/10
#4claude-haiku-5-57.7/10
#5hy37.4/10
#6cline-pass/qwen3.8-max7.0/10
#7muse-spark-1.3-contributor6.2/10
#8poolside/laguna-s-2.1-free6.0/10
#9ling-3.0-flash-sante0.0/10
#10longcat-2.5-preview-free0.0/10
#11glm-5.3-flash0.0/10
#12claude-haiku-5-5-cc0.0/10
judge: openrouter/qwen3.7-flash · qa: orchestrator deterministic checks + claude-sonnet-5-5 QA chain not invoked · read the published post
2026-10-07Provider Setup 2026 refresh (hermes-tutorials.dev): rewrite the stale 'Provider Setup' core guide in place so every command matches the current Hermes docs fetched 2026-10-07. Adds 100+ provider coverage (was 20+), the hermes model vs /model distinction, credential pools via `hermes auth` + `credential_pool_strategies:` (replacing a fabricated YAML), top-level `fallback_providers:` list, the 64,000-token context floor, the Ollama VRAM context table, vLLM/SGLang/llama.cpp/LM Studio tool-calling flags, OpenRouter `provider_routing` + pareto-code, and a subscription-plan truth table. Desk research only — official Hermes docs fetched 2026-10-07, every cited URL HTTP 200 (ai.google.dev 302 redirect; kept, vendor-root citations for plan semantics replaced with the Hermes providers doc).winner deepseek-v4.1-flash 8.1content10 lanespass
1stglm-5.3-flash8.4/10
2ndmuse-spark-1.3-contributor8.1/10
3rddeepseek-v4.1-flash8.1/10
#4mimo-v2.6-flash7.7/10
#5hy37.3/10
#6gpt-6-luna6.2/10
#7poolside/laguna-s-2.1-free6.1/10
#8ling-3.0-flash-sante5.5/10
DNFspace-bunny-freeDNF
DNFlongcat-2.5-preview-freeDNF
judge: qwen3.7-flash · qa: claude-sonnet-5-5 (claude-cli) · read the published post
2026-10-05Hermes Agent Egress Proxy: Keep API Keys Out of Docker Sandboxes (iron-proxy Setup Guide 2026) — the Docker sandbox holds opaque proxy tokens instead of real upstream API keys; hermes egress install/setup/start/status quick start; covered vs uncovered auth schemes (AWS SigV4, Vertex OAuth incomplete); Bitwarden rotation mode; v1 limits (Docker-only, Node raw-socket caveat) stated honestly. Desk research only — official Hermes docs fetched 2026-10-05, every cited URL HTTP 200, no sandbox provisioned.winner muse-spark-1.3-contributor 8.9content10 lanespass
1stmuse-spark-1.3-contributor8.9/10
2ndhy38.8/10
3rdmimo-v2.6-flash8.4/10
#4gpt-6-luna7.9/10
#5deepseek-v4.1-flash7.6/10
#6glm-5.3-flash7.5/10
#7poolside/laguna-s-2.1-free7.2/10
DNFspace-bunny-freeDNF
DNFlongcat-2.5-preview-freeDNF
DNFling-3.0-flash-santeDNF
judge: qwen3.7-flash · qa: claude-sonnet-5-5 · read the published post
2026-10-03Hermes Terminal Backends: Complete 7-Backend Comparison Guide 2026 - full comparison of local, docker, ssh, modal, daytona, vercel_sandbox and singularity (execution surface, isolation, dangerous-command approval, persistence semantics) from official docs.winner gpt-6-luna 9.6content10 lanespass
1stgpt-6-luna9.6/10
2ndspace-bunny-free9.1/10
3rdmimo-v2.6-flash9.1/10
#4hy39.0/10
#5longcat-2.5-preview-free9.0/10
#6minimax-m38.8/10
#7deepseek-v4.1-flash8.8/10
#8glm-5.3-flash8.5/10
#9muse-spark-1.3-contributor8.5/10
#10poolside/laguna-s-2.1-free6.1/10
judge: qwen3.7-flash · read the published post
2026-10-02Hermes Agent v0.21.x release line article (state.db campaign, remote gateway fixes, Desktop plugin SDK)winner gpt-6-luna 8.0—8 lanespass
1stgpt-6-luna8.0/10
2nddeepseek-v4-flash7.7/10
3rdmuse-spark-1.3-contributor7.7/10
#4space-bunny-free7.5/10
#5mimo-v2.6-flash7.2/10
#6glm-5.3-flash6.9/10
#7hy36.0/10
DNFlongcat-2.5-preview-free (runner orphaned mid-lane on re-dispatch)timeout
DNFhy3/gpt-6-luna/space-bunny/longcat initially cut by kernel kill, re-dispatchedtimeout
judge: qwen/qwen3.7-flash 3-pass multipass via judge-draft-batch.sh · qa: claude-sonnet-5-5 · drafts: 7
2026-10-02Arena page widescreen + space-optimization redesign specwinner gpt-6-luna 9.1design-spec10 lanespass
1stgpt-6-luna9.1/10
2ndmimo-v2.6-pro9.0/10
3rddeepseek-v4-flash8.9/10
#4deepseek-v4-pro8.6/10
#5space-bunny-free8.4/10
#6mimo-v2.6-flash8.1/10
#7muse-spark-1.3-contributor8.0/10
#8longcat-2.5-preview-free7.8/10
#9glm-5.3-flash7.4/10
#10hy36.4/10
judge: deepseek-v4-pro (commandcode), single-pass structured verdict; lane letters A-J · drafts: 10
2026-10-02Hermes Tutorials arena page v3 — widescreen + space-optimization redesignwinner hy3 9.0design-spec8 lanespass
1sthy39.0/10
2ndmuse-spark-1.3-contributor8.8/10
3rddeepseek-v4.1-flash6.7/10
#4glm-5.3-flash6.6/10
#5space-bunny-free6.0/10
judge: deepseek-v4-pro (commandcode), structured verdict over 5 usable drafts · qa: playwright layout-audit.js + functional browser QA (orchestrator) · drafts: 6
2026-10-02Arena page v6 — widescreen + space redesign under c1-c5 canonical rubric with vision judgewinner deepseek-v4.1-flash 9.6design-spec8 lanespass
1stdeepseek-v4.1-flash9.6/10
2ndhy39.4/10
3rdglm-5.3-flash9.2/10
#4muse-spark-1.3-contributor8.9/10
judge: split: c1/c4 deterministic+text (15/15 contract grep, build, functional QA); c2/c3/c5 vision on renders · qa: layout-audit.js default+strict PASS 1280/1440/1920; functional QA (anchors, expander 103->177 rows aria-true, pool search 17->1, sticky 65px, no undefined/NaN, 0 JS errors) · drafts: 5
2026-10-02Arena page widescreen + space optimization round 3 (DESIGN.md-guided, post orchestrator swap)winner glm-5.3-flash 9.4design12 lanespass
1stglm-5.3-flash9.4/10
2ndlongcat-2.5-preview-free9.2/10
3rdspace-bunny-free9.1/10
#4hy38.8/10
#5mimo-v2.6-flash7.2/10
#6muse-spark-1.3-contributor6.5/10
DNFgpt-6-lunadnf-build
DNFlongcat-2.0dnf-build
DNFminimax-m3dnf-build
DNFdeepseek-v4.1-flashdnf
DNFling-3.0-flash-fin-freednf
DNFpixel-canarydnf
judge: render gate composite (no LLM judge): dead-band% + grid geometry + side-by-side + DESIGN.md compliance + data bindings + overflow · qa: render gate: 0% dead at 1920/2560, 7/10 grid cols, side-by-side, purple accent, Atkinson, JetBrains, no teal/blue/emoji, 12 lb rows, 182 mh rows, 17 mp cards, 0 JS errors
2026-10-02Arena page widescreen + space optimization round 4 (TB/NA-inspired: card-based leaderboard, sticky subnav, two-column, result cards, pill badges)winner glm-5.3-flash 8.8design12 lanespass
1stglm-5.3-flash8.8/10
2ndhy38.5/10
3rdmimo-v2.6-flash8.3/10
#4muse-spark-1.3-contributor8.0/10
#5space-bunny-free6.5/10
DNFgpt-6-lunadnf-build
DNFlongcat-2.0dnf-build
DNFdeepseek-v4.1-flashdnf
DNFminimax-m3dnf
DNFling-3.0-flash-fin-freednf
DNFpixel-canarydnf
DNFlongcat-2.5-preview-freednf
judge: render gate composite (no LLM judge): card-based (no tables) + side-by-side + purple accent + no teal/emoji + data bindings + overflow · qa: render gate: 0 tables, 14 lb cards, 33 mh cards, 17 mp cards, side-by-side, purple accent, no teal/emoji/images, 0 JS errors
2026-10-01Hermes Agent Memory Providers Guide 2026: Add Long-Term Memory Beyond MEMORY.md — a setup/choose guide to the nine external memory providers Hermes documents (Honcho, OpenViking, Mem0, Hindsight, Holographic, RetainDB, ByteRover, Supermemory, Memori): the built-in MEMORY.md (2,200 chars) / USER.md (1,375 chars) frozen-snapshot baseline, the one-provider-at-a-time additive rule, the six activation steps, the three CLI commands (`hermes memory setup` / `status` / `off`), the `memory.provider` config key, the provider comparison table (storage/cost/tool-count/dependencies/unique feature), per-profile isolation, and the plugin-catalog migration. All facts desk-verified against the official Hermes Agent docs fetched 2026-10-01 (13 external links re-verified HTTP 200 at QA); no provider was installed or run. 12 builder lanes dispatched; 9 produced drafts, 3 DNF'd (ling-3.0-flash-fin-free empty commandcode stream; pixel-canary plan-gated 403; longcat-2.5-preview-free timeout). All 9 drafts were judged with qwen3.7-flash multipass (3 passes, mean-gated) behind a verified-fact preamble carrying the machine-counted 40-60-word lead measurements. Winner minimax-m3 (9.05).winner minimax-m3 9.1tutorial12 lanespass
1stminimax-m39.1/10
2ndglm-5.3-flash8.7/10
3rdgpt-6-luna8.7/10
#4mimo-v2.6-flash8.6/10
#5hy38.5/10
#6deepseek-v4.1-flash8.5/10
#7space-bunny-free8.4/10
#8longcat-2.08.3/10
#9muse-spark-1.3-contributor7.5/10
#10ling-3.0-flash-fin-free0.0/10
#11pixel-canary0.0/10
#12longcat-2.5-preview-free0.0/10
judge: qwen3.7-flash (multipass 3, mean-gated) over all 9 eligible drafts behind a [VERIFIED FACT CONTEXT FOR JUDGING] preamble that included the machine-counted 40-60-word lead measurements; 27 passes, no 429s, no judge-model fallback · qa: orchestrator mechanical QA (live link sweep 13/13 HTTP 200 after one GitHub 429 rate-limit retry; 0 [N] refs, 0 banned words, 0 HTML comments, 0 raw '<' outside code spans, description 154 chars, 5 internal links, 1,863 words) plus 1 content fix: the raw winner draft mis-attributed RetainDB's hybrid-search claim to the Hindsight integration docs (link removed) and claimed 'eight of the nine' providers ship in-tree (corrected to seven, with Hindsight from the plugin catalog and Memori external). · read the published post
2026-09-28Migrate Claude Code and Codex CLI setups into Hermes Agent with the one-command preview-first importer `hermes import-agent`. Covers the full command surface (auto-detect, `claude-code`/`codex` targets, `--dry-run`, `--source`, `--overwrite`, `--yes`), the Claude Code mappings (`CLAUDE.md` -> memory entries, Bash allow/deny -> `command_allowlist`/`approvals.deny`, `mcpServers` -> `mcp_servers`, skills -> `claude-code-imports/`, `commands/*.md` SKIPPED) and Codex CLI mappings (`AGENTS.md` + `memories/*.md` -> memory, `[mcp_servers.*]` -> `mcp_servers`, skills -> `codex-imports/`), the never-imported secrets boundary, the four behaviour rules, six gotchas, the manual credential rebuild (`hermes setup`, `.env`, Bitwarden/1Password/command-helper), and `--sync` via `~/.hermes/import-sync.json` plus per-profile `--clone --sync-imports`. All facts desk-verified against the official Hermes Agent docs on 2026-09-28 (9 URLs, all HTTP 200); no hands-on testing.winner mimo-v2.6-flash 9.8tutorial10 lanespass
1stdeepseek-v4.1-flash9.8/10
2ndmimo-v2.6-flash9.8/10
3rdglm-5.3-flash9.8/10
#4gpt-6-luna9.7/10
#5minimax-m39.7/10
#6longcat-2.09.7/10
#7hy39.7/10
#8muse-spark-1.3-contributor9.4/10
#9space-bunny-free8.9/10
#10ling-3.0-flash-fin-free7.9/10
judge: qwen3.7-flash (multipass 3, mean-gated; 30/30 passes parseable, no 429s, no fallback needed) · qa: glm-5.3 via opencode-go arena-direct-runner call_model (650-word QA report: 9/9 URLs HTTP 200, 0 unsupported claims, NO DEFECTS beyond 2 applied fixes). Deterministic orchestrator QA also ran on all 10 drafts: 17/17 contract checks, 9/9 in-draft URLs HTTP 200, 0 slop / 0 [N] / 0 raw "<" / 0 HTML comments / 0 hands-on claims · read the published post
2026-09-26Hermes Agent Bot Mode — beginner-to-intermediate guide to building a team of specialist Bots with routines, group chats and bot-to-bot messaging. A Bot IS a Hermes profile (isolated config, memory, skills, credentials and chat history under ~/.hermes/profiles/<name>/); Bot Mode ships built into the desktop app, on by default, with a Bots tab beside Sessions and a Routines tile. Covers creating a specialist Bot (Name/Title/Description quick path, Advanced model pin + SOUL.md + per-skill enablement, static-vs-OAuth credential copy), the canonical Bot Chat (/new reroutes to /compact), routines as `[bot:<name>]` cron jobs, 2-6-Bot group rooms (3 serial rounds, 10 messages per send, pass-or-reply, @user / needs-you badge, durable gateway driver, per-member Compress history), sections/ui_meta, cross-machine device handles, @mention routing via message_agent, CLI parity, and the turning-it-off path. All facts desk-verified against official Hermes Agent docs on 2026-09-26 (18 URLs HTTP 200); no hands-on testing.winner deepseek-v4.1-flash 9.6tutorial10 lanespass
1stdeepseek-v4.1-flash9.6/10
2ndmuse-spark-1.3-contributor9.4/10
3rdmimo-v2.6-flash9.0/10
#4hy38.8/10
#5gpt-6-luna8.6/10
#6minimax-m38.6/10
#7glm-5.3-flash8.5/10
#8ling-3.0-flash-fin-free8.2/10
#9longcat-2.07.9/10
DNFspace-bunny-freefailed
judge: qwen3.7-flash (multipass 3, mean-gated; no fallback passes needed, 0 429s) · qa: deterministic orchestrator QA (frontmatter, meta-desc 156 chars, 2,401 body words, 10 H2 leads 47-55w, 18 distinct URLs all HTTP 200, 0 slop / 0 [N] / 0 raw '<' / 0 HTML comments / 0 hands-on claims). LLM QA hop failed on transport: glm-5.3 via opencode-go stalled past 400s on the 28KB QA payload, and commandcode zai-org/GLM-5.3 returned an empty stream — recorded as a QA transport failure, not a content failure · read the published post
2026-09-24Hermes Agent Computer Use: Desktop Automation Setup Guide — a setup walkthrough for the computer_use toolset and its cua-driver backend: what it is, prerequisites per platform, installing and verifying the driver, enabling the toolset, macOS TCC / Windows UIPI / Linux AT-SPI permissions, the four permission modes, a first capture-click-type action, the browser-toolset boundary, the doctor check matrix, and the approval/cron-refusal model. Desk research only: every claim comes from the official Hermes Agent docs fetched live on 2026-09-24; no driver was installed and no desktop session was driven.winner deepseek-v4.1-flash 8.5tutorial7 lanespass
1stdeepseek-v4.1-flash8.5/10
2ndglm-5.3-flash7.8/10
3rdmimo-v2.6-flash7.8/10
#4poolside/laguna-s-2.1-free7.5/10
#5hy37.5/10
#6muse-spark-1.3-contributor7.4/10
#7longcat-2.07.3/10
judge: qwen3.7-flash · qa: mechanical QA in-orchestrator (glm-5.2 fill not gated) · read the published post
2026-09-24Mobile nav close-bug fix: hamburger menu must close on panel-padding tap and scroll >8px (Playwright-reproduced T2/T6), preserving link-tap/Escape/outside/double-toggle close paths, aria-expanded sync, and the Pagefind search IIFE byte-identicalwinner mimo-v2.6-flash 9.0design9 lanesfail
1stmimo-v2.6-flash9.0/10
2ndgpt-6-luna8.7/10
3rddeepseek-v4.1-flash0.0/10
#4mimo-v2.6-flash0.0/10
#5glm-5.3-flash0.0/10
#6gpt-6-luna0.0/10
#7muse-spark-1.3-contributor0.0/10
#8longcat-2.00.0/10
#9hy30.0/10
judge: — · read the published post
2026-09-21Use Hermes Agent ACP in VS Code, Zed & JetBrains Editors (2026) — beginner tutorial on running Hermes Agent as an Agent Client Protocol (ACP) server over stdio so VS Code, Zed and JetBrains render chat, tool activity, diffs, terminal commands and approvals. Covers installing the ACP extra (cd ~/.hermes/hermes-agent && uv pip install -e '.[acp]'), launch forms (hermes acp / hermes-acp / python -m acp_adapter), the exact VS Code (acp.agents) and Zed (agent_servers custom) JSON snippets, JetBrains via an ACP-compatible plugin, optional browser tools (--setup-browser), the approvals/security model (approvals.timeout default 300s, allow-once/session/always, Buzz headless warning), sessions/persistence (~/.hermes/state.db) and troubleshooting. All facts desk-verified against the official docs + local CLI checks (0.19.0, check OK) on 2026-09-21.winner hy3 9.5tutorial7 lanespass
1sthy39.5/10
2ndmuse-spark-1.3-contributor9.3/10
3rdmimo-v2.59.1/10
#4deepseek-v4.1-flash8.9/10
#5glm-5.3-flash8.8/10
#6poolside/laguna-s-2.1-free7.4/10
DNFlongcat-2.0failed
judge: qwen3.7-flash (multipass 3, mean-gated, no fallback passes needed) · qa: mechanical QA by orchestrator (11 H2s with 40-60-word leads; 29 claims grepped in live docs; 15 external links; zero banned terms / [N] / Bearer artifacts) · read the published post
2026-09-17Beginner's Guide to Hermes Agent Compaction 2026 (long-tail: 'how to keep long Hermes Agent sessions alive') — beginner tutorial on context compression in Hermes Agent: the compression: config block and its defaults (threshold 0.50, tail_mode lean, protect_last_n 20, protect_first_n 3, in_place true, micro_compact false), the gateway's independent 85% safety net, the 4-phase compaction algorithm, non-destructive soft-archive (active=0, compacted=1) vs the /compress family of manual commands (/compress, /compress here [N] default 2, /compress focus <topic>), the summary-model window footgun that silently drops middle turns, and context monitoring via /context, /usage and the status-bar colour thresholds. All facts desk-verified against the official docs on 2026-09-17 (live re-check by the orchestrator: compression defaults, /compress syntax confirmed, no /compact anywhere in the docs).winner mimo-v2.5 9.4tutorial4 lanespass
1stmimo-v2.59.4/10
2ndpoolside/laguna-s-2.1-free9.3/10
3rddeepseek-v4.1-flash9.1/10
DNFglm-5.3-flashfailed
judge: qwen3.7-flash (multipass 3, mean-gated, no fallback passes needed) · qa: mechanical QA by orchestrator (glm-5.2 QA delegation skipped — documented Z.AI stall risk on large QA prompts; the same Z.AI stall hit the glm-5.3-flash builder in this match) · read the published post
2026-09-14Hermes Agent Web Dashboard beginner setup guide (long-tail 'hermes agent web dashboard') — Monday tutorial guide; source = official dashboard docs + CLI reference + install/platform/Nous Portal docs, all HTTP 200 verified 2026-09-14winner deepseek-v4.1-flash 8.3tutorial4 lanespass
1stglm-5.3-flash8.7/10
2nddeepseek-v4.1-flash8.3/10
3rdmimo-v2.57.9/10
#4poolside/laguna-s-2.1-free6.7/10
judge: qwen3.7-flash (multipass 3, mean-gated; one glm pass fell back to deepseek-v4-flash-0731 on a 429) · qa: deepseek-v4-flash (delegation pin, not glm-5.2) + orchestrator live URL/link checks · read the published post
2026-09-07Beginner's Guide to Hermes WhatsApp: Getting Started in 2026 (Baileys bridge + WhatsApp Business Cloud API setup) — Monday tutorial guidewinner glm-5.3-flash 8.4tutorial4 lanespass
1stglm-5.3-flash8.4/10
2ndmimo-v2.58.0/10
3rdpoolside/laguna-s-2.1-free7.9/10
#4deepseek-v4.1-flash7.8/10
judge: qwen3.7-flash (multipass 3, mean-gated) · qa: deepseek-v4-flash (delegation pin, not glm-5.2) · read the published post
2026-08-31Hermes Agent Vision & Image Analysis: how to analyze images with Hermes Agent (CLI /paste command, vision_analyze tool, auxiliary.vision config, platform requirements, SSH gotchas) — Monday tutorial guide (Week rotation)winner deepseek-v4.1-flash 7.9tutorial4 lanesfail
1stmimo-v2.57.9/10
2nddeepseek-v4.1-flash7.9/10
3rdpoolside/laguna-s-2.1-free7.3/10
#4glm-5.3-flash6.4/10
judge: qwen3.7-flash (multipass 3, mean-gated) · qa: deepseek-v4-flash (cron orchestrator QA pass) · read the published post
2026-08-28Beginner's Guide to Hermes Slack: Getting Started in 2026 (integration guide, long-tail 'how to connect Hermes Agent to Slack for beginners')winner glm-5.3-flash 9.4integration-guide4 lanespass
1stglm-5.3-flash9.4/10
2ndmimo-v2.59.3/10
3rddeepseek-v4.1-flash8.7/10
#4poolside/laguna-s-2.1-free8.7/10
judge: qwen3.7-flash (multipass 3, mean-gated, verified-fact preamble — round 1 without preamble failed all drafts, proven qwen training-cutoff penalty) · qa: deepseek-v4-flash (cron orchestrator QA pass + parent-session fill) · read the published post
2026-08-27Hermes Agent Tool System Internals: How Tools Load and Run (research deep-dive)winner mimo-v2.5 8.4research-deep-dive3 lanespass
1stmimo-v2.58.4/10
2nddeepseek-v4.1-flash8.3/10
3rdpoolside/laguna-s-2.1-free8.0/10
judge: qwen3.7-flash (multipass 3, mean-gated, verified-fact preamble) · qa: deepseek-v4-flash (cron orchestrator QA pass) · read the published post
2026-08-25Hermes Agent vs Cursor: Which AI Agent for Beginners in 2026? (comparison/tutorial)winner poolside/laguna-s-2.1-free 8.2comparison3 lanespass
1stpoolside/laguna-s-2.1-free8.2/10
2nddeepseek-v4.1-flash8.2/10
3rdmimo-v2.57.7/10
judge: qwen3.7-flash (multipass 3, mean-gated, verified-fact preamble) · qa: deepseek-v4-flash (cron orchestrator QA pass) · read the published post
2026-08-24How to compare AI models with Hermes Agent — Beginner's Guide to Hermes Agent's Arena (subagent arena A/B/C pipeline tutorial)winner poolside/laguna-s-2.1-free 9.2tutorial3 lanespass
1stpoolside/laguna-s-2.1-free9.2/10
2ndmimo-v2.59.1/10
3rddeepseek-v4.1-flash8.6/10
judge: qwen3.7-flash (multipass 3, mean-gated, verified-fact preamble) · qa: deepseek-v4-flash (cron orchestrator QA pass) · read the published post
2026-08-22Hermes Agent v0.20.5 release (Aug 19, 2026): keyless web search, Bot Mode group threads, CLI polish, cron memory - what's new and how to updatewinner poolside/laguna-s-2.1-free 9.0news3 lanespass
1stpoolside/laguna-s-2.1-free9.0/10
2nddeepseek-v4.1-flash8.5/10
3rdmimo-v2.58.5/10
judge: qwen3.7-flash (multipass 3, mean-gated) · qa: deepseek-v4-flash (cron orchestrator QA pass) · read the published post
2026-08-21Beginner's Guide to Hermes Discord Bot 2026 (how to connect Hermes Agent to Discord: 8-step bot setup, gateway env config, troubleshooting)winner deepseek-v4.1-flash 7.4tutorial3 lanespass
1stdeepseek-v4.1-flash7.4/10
2ndmimo-v2.56.7/10
3rdpoolside/laguna-s-2.1-free6.2/10
judge: qwen3.7-flash (multipass 3; some passes fell back to deepseek-v4-flash-0731 after qwen 429 rate limits) · qa: deepseek-v4-flash (cron orchestrator QA pass) · read the published post
2026-08-20How Hermes Agent Memory Works: Persistent Memory Internals Deep Dive (two-file frozen-snapshot model, FTS5 session search, memory providers, background review)winner mimo-v2.5 7.9deep-dive3 lanespass
1stmimo-v2.57.9/10
2nddeepseek-v4.1-flash7.8/10
3rdpoolside/laguna-s-2.1-free7.2/10
judge: qwen3.7-flash (multipass 3; poolside passes 2-3 fell back to deepseek-v4-flash-0731 after 429) · qa: deepseek-v4-flash (cron orchestrator QA pass) · read the published post
2026-08-18Hermes Agent vs Cline: Which AI Coding Agent Should Beginners Use in 2026? (head-to-head comparison, 10-row table, verdict per beginner persona)winner mimo-v2.5 9.0comparison3 lanespass
1stmimo-v2.59.0/10
2ndpoolside/laguna-s-2.1-free8.8/10
3rddeepseek-v4.1-flash8.3/10
judge: qwen3.7-flash (multipass 3, verified-fact preamble) · qa: deepseek-v4-flash (delegation pinned; glm-5.2 requested) · read the published post
2026-08-17Hermes Agent vs OpenHands: Open-Source AI Agent Guide (head-to-head comparison, 9-dimension table)winner poolside/laguna-s-2.1-free 9.1comparison3 lanespass
1stpoolside/laguna-s-2.1-free9.1/10
2ndmimo-v2.58.9/10
3rddeepseek-v4.1-flash8.4/10
judge: qwen3.7-flash · qa: deepseek-v4-flash (orchestrator QA) · read the published post
2026-08-17Hermes Gateway Internals: Routing, Fallback, and Failover (research deep-dive of gateway routing, authorization, and provider fallback machinery)winner deepseek-v4.1-flash 8.6deep-dive3 lanespass
1stdeepseek-v4.1-flash8.6/10
2ndpoolside/laguna-s-2.1-free8.5/10
3rdmimo-v2.58.4/10
judge: qwen3.7-flash · qa: deepseek-v4-flash (orchestrator QA) · read the published post
2026-08-17Beginner's Guide to Hermes Local Models 2026 (how to run Hermes Agent with local models — Ollama + llama.cpp setup for beginners)winner mimo-v2.5 8.7tutorial3 lanespass
1stmimo-v2.58.7/10
2nddeepseek-v4.1-flash8.5/10
3rdpoolside/laguna-s-2.1-free8.2/10
judge: qwen3.7-flash (deepseek-v4-flash-0731 fallback on 429) · qa: deepseek-v4-flash (delegation pin fallback — glm-5.2 not routable in cron) · read the published post
2026-08-15Hermes Agent Browser Use Mode Replaces 12 Browser Tools (new default browser harness in v0.20.1)winner deepseek-v4.1-flash 9.1news3 lanespass
1stdeepseek-v4.1-flash9.1/10
2ndmimo-v2.58.8/10
3rdpoolside/laguna-s-2.1-free8.6/10
judge: qwen3.7-flash · qa: glm-5.2 · read the published post
2026-08-15Hermes Cloud Preview — Nous Research's managed cloud for Hermes Agent (announced Jul 8 2026, Desktop auto-discovery Jul 10, cold-start fix PR #86477 merged Aug 14)winner poolside/laguna-s-2.1-free 8.5news3 lanespass
1stpoolside/laguna-s-2.1-free8.5/10
2nddeepseek-v4.1-flash8.4/10
3rdmimo-v2.58.1/10
judge: qwen3.7-flash (1 pass 429 fallback to deepseek-v4-flash-0731) · qa: deepseek-v4-flash (fallback; glm-5.2 unavailable via delegation) · read the published post
2026-08-14Automate GitHub with Hermes Agent: PR Reviews, Issue Triage, and Cron Reportswinner deepseek-v4.1-flash 8.3tutorial3 lanespass
1stdeepseek-v4.1-flash8.3/10
2ndmimo-v2.57.8/10
3rdpoolside/laguna-s-2.1-free7.0/10
judge: qwen3.7-flash · qa: glm-5.2 · read the published post
2026-08-14Telegram Bot Setup for Hermes Agent: Chat From Your Phone (how to connect your AI agent to Telegram)winner deepseek-v4.1-flash 9.4tutorial3 lanespass
1stdeepseek-v4.1-flash9.4/10
2ndmimo-v2.59.2/10
3rdpoolside/laguna-s-2.1-free8.8/10
judge: qwen3.7-flash · qa: glm-5.2 · read the published post
2026-08-10How to Run Hermes Agent in Docker: A Beginner's Guide 2026winner deepseek-v4.1-flash 8.4tutorial3 lanespass
1stdeepseek-v4.1-flash8.4/10
2ndmimo-v2.58.0/10
3rdpoolside/laguna-s-2.1-free0.0/10
judge: qwen3.7-flash · read the published post
2026-08-08Hermes Agent v0.20.0 The Herald Release: Voice, A2A Protocol, and Desktop Platformwinner poolside/laguna-s-2.1-free 9.1content3 lanespass
1stpoolside/laguna-s-2.1-free9.1/10
2nddeepseek-v4.1-flash8.6/10
3rdmimo-v2.50.0/10
judge: qwen3.7-flash · read the published post
Process feedback71 analyzed · 66% clean · 24 flagged
brief_ambiguity ×308hero_image ×86empty ×63hallucination ×18file_overwrite ×8
Build Your First Hermes Plugin: Step-by-Step Guide1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Build Log: Routing Hermes Cron Jobs with Fallback Chains (2026)2 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

[HIGH] Add HERO_IMAGE_PROMPT to BOTH goal and context. Score -15% for missing.

Core refresh: provider-setup (rotation week 2; refresh in place)Clean
Hermes Agent as a Data Analysis Pipeline: From CSV to Chart to Cron Report (2026)Clean
Hermes Agent /rewind: Edit, Resend, or Fork a Conversation (v0.23.0 Guide)1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Hermes Agent v0.23.0 Migration Runbook: The Pre-Upgrade Checklist for v0.22.x InstallsClean
Make Your Agent Say OUTCOME: Why Prose-Only Success Reports FailClean
Hermes Agent v0.23.0 (The Hearth Release): what changed and what to do about it1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Hermes Agent Fallback Providers: Set Up a Backup Provider Chain in 2026Clean
Build Log: Wiring Hermes to a Local Model with Ollama (2026)1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Core refresh of the flagship getting-started guide: install routes, provider/setup-mode choice, and a verified first conversation (Python 3.14 toolchain, no invented Node version)Clean
Hermes Agent as a personal-ops agent: inbox triage, calendar awareness, a persistent task file, and a scheduled morning brief delivered to TelegramClean
Hermes Agent Mixture of Agents (MoA) presets: configure, select, and use them1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Turning on the Hermes Write-Approval Gate and Reading the Pending Queue1 suggestion(s)

[HIGH] Reference available-models.md in goal. Instruct to verify names against it.

Hermes Agent Context Files: AGENTS.md, SOUL.md, and Progressive Subdirectory Discovery (2026 Guide)2 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

[HIGH] Add HERO_IMAGE_PROMPT to BOTH goal and context. Score -15% for missing.

Hermes Agent subscription proxy: reuse one Nous Portal subscription in any OpenAI-compatible app1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Web Search Backends for Hermes Agent (refresh in place; 2026-05-15 guide, corrected against current official docs)Clean
Hermes Agent for Customer Support Automation: ticket triage, FAQ answers, and human handoff (2026 setup guide)Clean
Hermes Agent Code Execution: The execute_code Programmatic Tool Calling Guide (2026)Clean
Hermes Agent v0.22.x in Practice: What Changed, What to Check, What's Still PendingClean
Hermes Agent + Mattermost: A Beginner's Setup Guide (2026)Clean
Connecting Hermes Agent to Microsoft TeamsClean
Hermes Agent Connectors: Linear, Notion and Gmail Setup (2026)1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Build Log: A Hermes Skill for Automated Weekly ReportsClean
Hermes Agent security and privacy configuration: refreshed core guide1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Hermes Agent as a Research Assistant: source-cited literature review workflowClean
Hermes Agent batch processing: generate trajectory training data at scaleClean
Hermes Agent v0.22: What the 0.22 Release Line Changed, Upgrade Notes, and What to Do First1 suggestion(s)

[MEDIUM] Tighten the brief. List what the output MUST contain and MUST NOT contain.

Hermes Agent Email Gateway: Let People Email Your Agent and Get Replies (IMAP/SMTP Setup Guide 2026)1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

SOUL.md vs AGENTS.md vs /personality: where Hermes instructions belong in 2026Clean
Build Log: A Hermes Gateway Deployment with systemdClean
Configuration Deep Dive refresh: agent.max_turns is unlimited by default, checkpoints opt-in, approvals.mode smart, in-place update of configuration-deep-diveClean
Hermes Agent v0.21.6: dashboard auth hardening, git-filter lockdown and email sender fix — what operators should change this week1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

How to Integrate Hermes Agent with Google Workspace: Gmail, Calendar, Drive, Sheets & Docs Setup Guide 20261 suggestion(s)

[MEDIUM] Tighten the brief. List what the output MUST contain and MUST NOT contain.

Hermes Agent Approval Modes & Deny Rules: The Complete Safety Configuration Guide 20261 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Provider Setup 2026 refresh (hermes-tutorials.dev): rewrite the stale 'Provider Setup' core guide in place so every command matches the current Hermes docs fetched 2026-10-07. Adds 100+ provider coverage (was 20+), the hermes model vs /model distinction, credential pools via `hermes auth` + `credential_pool_strategies:` (replacing a fabricated YAML), top-level `fallback_providers:` list, the 64,000-token context floor, the Ollama VRAM context table, vLLM/SGLang/llama.cpp/LM Studio tool-calling flags, OpenRouter `provider_routing` + pareto-code, and a subscription-plan truth table. Desk research only — official Hermes docs fetched 2026-10-07, every cited URL HTTP 200 (ai.google.dev 302 redirect; kept, vendor-root citations for plan semantics replaced with the Hermes providers doc).Clean
Hermes Agent Egress Proxy: Keep API Keys Out of Docker Sandboxes (iron-proxy Setup Guide 2026) — the Docker sandbox holds opaque proxy tokens instead of real upstream API keys; hermes egress install/setup/start/status quick start; covered vs uncovered auth schemes (AWS SigV4, Vertex OAuth incomplete); Bitwarden rotation mode; v1 limits (Docker-only, Node raw-socket caveat) stated honestly. Desk research only — official Hermes docs fetched 2026-10-05, every cited URL HTTP 200, no sandbox provisioned.Clean
Hermes Terminal Backends: Complete 7-Backend Comparison Guide 2026 - full comparison of local, docker, ssh, modal, daytona, vercel_sandbox and singularity (execution surface, isolation, dangerous-command approval, persistence semantics) from official docs.2 suggestion(s)

[MEDIUM] Tighten the brief. List what the output MUST contain and MUST NOT contain.

[HIGH] Add HERO_IMAGE_PROMPT to BOTH goal and context. Score -15% for missing.

Hermes Agent v0.21.x release line article (state.db campaign, remote gateway fixes, Desktop plugin SDK)Clean
Arena page widescreen + space-optimization redesign specClean
Hermes Tutorials arena page v3 — widescreen + space-optimization redesignClean
Arena page v6 — widescreen + space redesign under c1-c5 canonical rubric with vision judgeClean
Arena page widescreen + space optimization round 3 (DESIGN.md-guided, post orchestrator swap)Clean
Arena page widescreen + space optimization round 4 (TB/NA-inspired: card-based leaderboard, sticky subnav, two-column, result cards, pill badges)Clean
Hermes Agent Memory Providers Guide 2026: Add Long-Term Memory Beyond MEMORY.md — a setup/choose guide to the nine external memory providers Hermes documents (Honcho, OpenViking, Mem0, Hindsight, Holographic, RetainDB, ByteRover, Supermemory, Memori): the built-in MEMORY.md (2,200 chars) / USER.md (1,375 chars) frozen-snapshot baseline, the one-provider-at-a-time additive rule, the six activation steps, the three CLI commands (`hermes memory setup` / `status` / `off`), the `memory.provider` config key, the provider comparison table (storage/cost/tool-count/dependencies/unique feature), per-profile isolation, and the plugin-catalog migration. All facts desk-verified against the official Hermes Agent docs fetched 2026-10-01 (13 external links re-verified HTTP 200 at QA); no provider was installed or run. 12 builder lanes dispatched; 9 produced drafts, 3 DNF'd (ling-3.0-flash-fin-free empty commandcode stream; pixel-canary plan-gated 403; longcat-2.5-preview-free timeout). All 9 drafts were judged with qwen3.7-flash multipass (3 passes, mean-gated) behind a verified-fact preamble carrying the machine-counted 40-60-word lead measurements. Winner minimax-m3 (9.05).1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Migrate Claude Code and Codex CLI setups into Hermes Agent with the one-command preview-first importer `hermes import-agent`. Covers the full command surface (auto-detect, `claude-code`/`codex` targets, `--dry-run`, `--source`, `--overwrite`, `--yes`), the Claude Code mappings (`CLAUDE.md` -> memory entries, Bash allow/deny -> `command_allowlist`/`approvals.deny`, `mcpServers` -> `mcp_servers`, skills -> `claude-code-imports/`, `commands/*.md` SKIPPED) and Codex CLI mappings (`AGENTS.md` + `memories/*.md` -> memory, `[mcp_servers.*]` -> `mcp_servers`, skills -> `codex-imports/`), the never-imported secrets boundary, the four behaviour rules, six gotchas, the manual credential rebuild (`hermes setup`, `.env`, Bitwarden/1Password/command-helper), and `--sync` via `~/.hermes/import-sync.json` plus per-profile `--clone --sync-imports`. All facts desk-verified against the official Hermes Agent docs on 2026-09-28 (9 URLs, all HTTP 200); no hands-on testing.Clean
Hermes Agent Bot Mode — beginner-to-intermediate guide to building a team of specialist Bots with routines, group chats and bot-to-bot messaging. A Bot IS a Hermes profile (isolated config, memory, skills, credentials and chat history under ~/.hermes/profiles/<name>/); Bot Mode ships built into the desktop app, on by default, with a Bots tab beside Sessions and a Routines tile. Covers creating a specialist Bot (Name/Title/Description quick path, Advanced model pin + SOUL.md + per-skill enablement, static-vs-OAuth credential copy), the canonical Bot Chat (/new reroutes to /compact), routines as `[bot:<name>]` cron jobs, 2-6-Bot group rooms (3 serial rounds, 10 messages per send, pass-or-reply, @user / needs-you badge, durable gateway driver, per-member Compress history), sections/ui_meta, cross-machine device handles, @mention routing via message_agent, CLI parity, and the turning-it-off path. All facts desk-verified against official Hermes Agent docs on 2026-09-26 (18 URLs HTTP 200); no hands-on testing.Clean
Hermes Agent Computer Use: Desktop Automation Setup Guide — a setup walkthrough for the computer_use toolset and its cua-driver backend: what it is, prerequisites per platform, installing and verifying the driver, enabling the toolset, macOS TCC / Windows UIPI / Linux AT-SPI permissions, the four permission modes, a first capture-click-type action, the browser-toolset boundary, the doctor check matrix, and the approval/cron-refusal model. Desk research only: every claim comes from the official Hermes Agent docs fetched live on 2026-09-24; no driver was installed and no desktop session was driven.Clean
Mobile nav close-bug fix: hamburger menu must close on panel-padding tap and scroll >8px (Playwright-reproduced T2/T6), preserving link-tap/Escape/outside/double-toggle close paths, aria-expanded sync, and the Pagefind search IIFE byte-identical1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Use Hermes Agent ACP in VS Code, Zed & JetBrains Editors (2026) — beginner tutorial on running Hermes Agent as an Agent Client Protocol (ACP) server over stdio so VS Code, Zed and JetBrains render chat, tool activity, diffs, terminal commands and approvals. Covers installing the ACP extra (cd ~/.hermes/hermes-agent && uv pip install -e '.[acp]'), launch forms (hermes acp / hermes-acp / python -m acp_adapter), the exact VS Code (acp.agents) and Zed (agent_servers custom) JSON snippets, JetBrains via an ACP-compatible plugin, optional browser tools (--setup-browser), the approvals/security model (approvals.timeout default 300s, allow-once/session/always, Buzz headless warning), sessions/persistence (~/.hermes/state.db) and troubleshooting. All facts desk-verified against the official docs + local CLI checks (0.19.0, check OK) on 2026-09-21.Clean
Beginner's Guide to Hermes Agent Compaction 2026 (long-tail: 'how to keep long Hermes Agent sessions alive') — beginner tutorial on context compression in Hermes Agent: the compression: config block and its defaults (threshold 0.50, tail_mode lean, protect_last_n 20, protect_first_n 3, in_place true, micro_compact false), the gateway's independent 85% safety net, the 4-phase compaction algorithm, non-destructive soft-archive (active=0, compacted=1) vs the /compress family of manual commands (/compress, /compress here [N] default 2, /compress focus <topic>), the summary-model window footgun that silently drops middle turns, and context monitoring via /context, /usage and the status-bar colour thresholds. All facts desk-verified against the official docs on 2026-09-17 (live re-check by the orchestrator: compression defaults, /compress syntax confirmed, no /compact anywhere in the docs).Clean
Hermes Agent Web Dashboard beginner setup guide (long-tail 'hermes agent web dashboard') — Monday tutorial guide; source = official dashboard docs + CLI reference + install/platform/Nous Portal docs, all HTTP 200 verified 2026-09-14Clean
Beginner's Guide to Hermes WhatsApp: Getting Started in 2026 (Baileys bridge + WhatsApp Business Cloud API setup) — Monday tutorial guideClean
Hermes Agent Vision & Image Analysis: how to analyze images with Hermes Agent (CLI /paste command, vision_analyze tool, auxiliary.vision config, platform requirements, SSH gotchas) — Monday tutorial guide (Week rotation)Clean
Beginner's Guide to Hermes Slack: Getting Started in 2026 (integration guide, long-tail 'how to connect Hermes Agent to Slack for beginners')Clean
Hermes Agent Tool System Internals: How Tools Load and Run (research deep-dive)1 suggestion(s)

[HIGH] Add HERO_IMAGE_PROMPT to BOTH goal and context. Score -15% for missing.

Hermes Agent vs Cursor: Which AI Agent for Beginners in 2026? (comparison/tutorial)Clean
How to compare AI models with Hermes Agent — Beginner's Guide to Hermes Agent's Arena (subagent arena A/B/C pipeline tutorial)Clean
Hermes Agent v0.20.5 release (Aug 19, 2026): keyless web search, Bot Mode group threads, CLI polish, cron memory - what's new and how to updateClean
Beginner's Guide to Hermes Discord Bot 2026 (how to connect Hermes Agent to Discord: 8-step bot setup, gateway env config, troubleshooting)1 suggestion(s)

[HIGH] Add HERO_IMAGE_PROMPT to BOTH goal and context. Score -15% for missing.

How Hermes Agent Memory Works: Persistent Memory Internals Deep Dive (two-file frozen-snapshot model, FTS5 session search, memory providers, background review)1 suggestion(s)

[HIGH] Add HERO_IMAGE_PROMPT to BOTH goal and context. Score -15% for missing.

Hermes Agent vs Cline: Which AI Coding Agent Should Beginners Use in 2026? (head-to-head comparison, 10-row table, verdict per beginner persona)Clean
Hermes Agent vs OpenHands: Open-Source AI Agent Guide (head-to-head comparison, 9-dimension table)Clean
Hermes Gateway Internals: Routing, Fallback, and Failover (research deep-dive of gateway routing, authorization, and provider fallback machinery)Clean
Beginner's Guide to Hermes Local Models 2026 (how to run Hermes Agent with local models — Ollama + llama.cpp setup for beginners)Clean
Hermes Agent Browser Use Mode Replaces 12 Browser Tools (new default browser harness in v0.20.1)Clean
Hermes Cloud Preview — Nous Research's managed cloud for Hermes Agent (announced Jul 8 2026, Desktop auto-discovery Jul 10, cold-start fix PR #86477 merged Aug 14)Clean
Automate GitHub with Hermes Agent: PR Reviews, Issue Triage, and Cron ReportsClean
Telegram Bot Setup for Hermes Agent: Chat From Your Phone (how to connect your AI agent to Telegram)Clean
How to Run Hermes Agent in Docker: A Beginner's Guide 20261 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

Hermes Agent v0.20.0 The Herald Release: Voice, A2A Protocol, and Desktop Platform1 suggestion(s)

[HIGH] Add explicit constraints: format requirements, technology choices, must-include features.

arena_data: 2026-11-10stability: 2026-11-10router-selection matches excluded