Is Hermes Agent v0.5.0 stable?
60/100Fact score: 60/100 · evidence-based, not a YES/NO verdict
“Fewer defensible rows than v0.4.0, but UI command dispatch, Matrix/Discord delivery, MCP args, provider streaming, and Telegram group behavior pass the gate.”
Stability Score
60Base
80
Evidence
-23
Votes
-0
Survived
+3
Curated
+0
80 base−23 evidence−0 votes+3 (3×1 days survived)=60
Affected components
uimatrixmcptelegramprovider-integrationmessage-deliverygatewaytool
Install this version
npm install hermes-agent@0.5.0Score Evidence
| Issue | Area | Type | Penalty | Reason |
|---|---|---|---|---|
| NousResearch/hermes-agent#3487 → | matrix | runtime | -3 | after v0.5.0 update, Matrix non-image file downloads and text MIME enrichment remained unresolved |
| NousResearch/hermes-agent#3577 → | provider-integration | tool-specific | -1 | 0.5.0 Claude Pro setup defaults to 1M context with no selectable context window |
| NousResearch/hermes-agent#3806 → | matrix | runtime | -3 | 0.5.0 Matrix E2EE encrypted file uploads are silently dropped due to missing callbacks |
| NousResearch/hermes-agent#3821 → | ui | runtime | -3 | v0.5.0 /model slash command is not wired into interactive CLI dispatch/autocomplete |
| NousResearch/hermes-agent#3947 → | mcp | runtime | -3 | v0.5.0 MCP bridge serializes numeric/array tool parameters as strings |
| NousResearch/hermes-agent#3979 → | telegram | runtime | -3 | v0.5.0 Telegram require_mention is ignored in groups and bot responds to every message |
| NousResearch/hermes-agent#4082 → | provider-integration | tool-specific | -1 | v0.5.0 Minimax auxiliary default selects a higher-cost model than expected |
| NousResearch/hermes-agent#22882 → | message-delivery | runtime | -3 | v2026.3.28 send_message Discord DMs fail with Unknown Channel when target is a user ID |
| NousResearch/hermes-agent#31063 → | provider-integration | runtime | -3 | 0.5.0 Gemini OpenAI-compatible streaming chunks with choices=0 can break conversation flow |