π Hermes Agent v0.9.0
"The Everywhere Release"
Released: April 13, 2026
Tag: v2026.4.13
Generated: April 14, 2026
π Release Metrics
63,281 lines added Β· 493 files changed Β· 5 days (April 8-13)
π― Strategic Positioning
From developer tool β Universal AI agent platform
- 16 messaging platforms β Telegram, Discord, Slack, WhatsApp, Signal, iMessage, WeChat, WeCom, and more
- Mobile-first β Native Android support via Termux
- Enterprise-ready β 16 security fixes, unified proxy support
- Chinese market β WeChat/WeCom for enterprise deployment
π± 16 Supported Platforms
NEW: iMessage (BlueBubbles), WeChat, WeCom
β¨ Headline Features (1/3)
π Local Web Dashboard
Browser-based management for Hermes Agent β configure settings, monitor sessions, browse skills, manage gateway without touching config files or terminal.
β‘ Fast Mode (/fast)
Priority processing for OpenAI and Anthropic models. Toggle /fast for significantly lower latency on GPT-5.4, Codex, and Claude models.
β¨ Headline Features (2/3)
π iMessage via BlueBubbles
- Full iMessage integration through BlueBubbles
- Auto-webhook registration, setup wizard
- Crash resilience and media support
π¨π³ WeChat & WeCom Support
- Native WeChat via iLink Bot API
- WeCom callback mode for enterprise apps
- Streaming, media uploads, markdown links
β¨ Headline Features (3/3)
π± Termux / Android Support
- Run Hermes natively on Android via Termux
- TUI optimizations for mobile screens
- Voice backend and /image command work on-device
ποΈ Background Process Monitoring
- watch_patterns parameter for real-time alerts
- Monitor errors, wait for events, watch build logs
- No polling required
π οΈ Developer Experience
- Native xAI & Xiaomi MiMo providers β First-class Grok and MiMo support
- Pluggable context engine β Custom context management via hermes plugins
- Debug toolkit β /debug command + hermes debug share for pastebin uploads
- Multi-arch Docker β amd64 + arm64, runs as non-root
- Bulk file sync β tar pipe for SSH/Modal backends
π Security Hardening (16 Fixes)
- CRITICAL: SMS RCE fix (Twilio webhook signature validation)
- Path traversal protection in checkpoint manager
- Shell injection neutralization in sandbox writes
- SSRF redirect guards in Slack image uploads
- API server auth enforcement
- Git argument injection prevention
- Approval button authorization
Enterprise Focus: Production-ready security posture with comprehensive hardening
ποΈ Infrastructure & Operations
Unified Proxy Support
- SOCKS proxy, DISCORD_PROXY, system proxy auto-detection
- Works behind corporate firewalls
Backup & Restore
- hermes backup β Full config snapshot
- hermes import β Restore or migrate between machines
- Includes sessions, skills, memory, configuration
π§ Tool System Improvements
- Unified spawn-per-call execution β Consistent environment handling
- Persistent sandbox environments β Survive between turns
- Vision auto-resize β Handles oversized images (20 MB limit)
- Voxtral TTS provider β Mistral AI voice synthesis
- TTS speed support β Edge TTS, OpenAI TTS, MiniMax
- MCP enhancements β --env and --preset support
π₯οΈ CLI & User Experience
- Native /model picker modal β Interactive provider β model selection
- Live per-tool elapsed timer β Restored in TUI spinner
- Stacked tool progress scrollback β Better visibility in TUI
- Random tips on session start β 279 tips across CLI + gateway
- hermes dump β Copy-pasteable setup summary for debugging
- WSL environment hint β Better Windows Subsystem for Linux support
π§ Context & Memory
- Improved context compression β Higher limits, tool tracking, degradation warnings
- /compress <focus> β Guided compression with focus topic
- Tiered context pressure warnings β Gateway deduplication
- Prevent agent from stopping mid-task β Compression floor, budget overhaul
- Propagate child activity to parent β During delegate_task
- Hindsight memory plugin β Feature parity, setup wizard
π€ Provider & Model Support
- xAI (Grok) β Native provider with direct API access
- Xiaomi MiMo β First-class provider, setup wizard, model catalog
- Qwen OAuth β Portal request support
- Structured API error classification β Smart failover decisions
- Rate limit header capture β Shown in /usage
- Fallback provider activation β On repeated empty responses
- OpenRouter variant tags β :free, :extended, :fast preserved
π‘ Gateway Platform Improvements
Discord
- Allowed channels whitelist config
- Forum channel topic inheritance
- DISCORD_REPLY_TO_MODE setting
Matrix
- Migrated from matrix-nio to mautrix-python
- SQLite crypto store (fixes E2EE decryption)
- Cross-signing recovery key verification
π Gateway Core Enhancements
- Inbound text batching β Discord, Matrix, WeCom + adaptive delay
- Surface natural mid-turn messages β Better chat platform integration
- WSL-aware gateway β Smart systemd detection
- Per-platform tool progress overrides β Customizable verbosity
- Configurable 'still working' interval β User-adjustable notifications
- /model switch persists β Across messages
- Drain in-flight work before restart β Graceful shutdown
π Migration Guide
Upgrade Steps
# Backup current installation
hermes backup
# Update Hermes Agent
cd ~/.hermes/hermes-agent
git pull
git checkout v2026.4.13
# Reinstall dependencies
pip install -e .
# Restart gateway (if running)
hermes gateway restart
# Verify installation
hermes doctor
β οΈ Breaking Changes & Deprecations
- Removed /prompt slash command β Prefix expansion footgun
- Removed dead LLM_MODEL env var β Migration clears stale entries
- Matrix migration β matrix-nio β mautrix-python (E2EE crypto store changes)
- Config priority change β config.yaml now takes priority over env vars for auxiliary settings
Note: Most changes are backward-compatible. Review release notes for platform-specific changes.
π Known Issues & Pain Points
- Matrix E2EE migration β Users may need to re-verify cross-signing keys
- BlueBubbles setup β Requires BlueBubbles server running on macOS
- WeChat/WeCom β Requires iLink Bot API key or WeCom enterprise app
- Termux limitations β Some features may have reduced functionality on Android
- Fast Mode availability β Only works with OpenAI Priority Processing and Anthropic fast tier
π₯ Community Response
GitHub Reactions: 29 total (18 π, 6 β€οΈ, 3 π, 2 π)
Key Sentiment Themes
- Platform expansion praised β iMessage and WeChat fill major gaps
- Security hardening welcomed β Enterprise users appreciate production focus
- Fast Mode excitement β Lower latency for interactive use cases
- Android support significant β Opens mobile developer market
- Web dashboard simplifies onboarding β Reduces CLI friction for new users
βοΈ Competitive Positioning
vs OpenClaw
- 16 platforms vs CLI-first β Hermes wins on deployment flexibility
- Mobile support β Hermes has Android, OpenClaw desktop-only
vs Claude Code
- Multi-platform vs single-platform β Hermes everywhere, Claude Code CLI-only
- Self-hosted vs cloud β Hermes offers full control
Chinese Market
- WeChat/WeCom β Enables enterprise deployment in China
- First-mover advantage β Few AI agents support Chinese platforms
β‘ Performance Impact
Fast Mode Latency Reduction
Significantly lower latency for OpenAI Priority Processing and Anthropic fast tier models
Context Compression Improvements
- Higher compression limits
- Better tool tracking
- Token-budget tail protection
Gateway Optimizations
- Inbound text batching reduces API calls
- Adaptive delay prevents rate limiting
πΊοΈ Roadmap Implications
- Platform expansion continues β More messaging platforms likely
- Mobile-first focus β iOS support may follow Android
- Enterprise features β Security and compliance will remain priority
- Chinese market penetration β WeChat/WeCom signal strategic focus
- Web dashboard evolution β May become primary interface for non-technical users
- Pluggable architecture β Context engine is first of many pluggable components
π¨βπ» Developer Impact
Positive
- More deployment options (16 platforms + Android)
- Better debugging tools (/debug, hermes dump)
- Easier onboarding (web dashboard)
- Production-ready security
Considerations
- Platform-specific setup complexity (BlueBubbles, WeChat)
- Matrix E2EE migration requires manual steps
- Fast Mode requires specific provider tiers
π Platform & Market Impact
- Democratizes AI agents β 16 platforms make Hermes accessible everywhere
- Mobile expansion β Android support opens new user segments
- Enterprise readiness β Security hardening enables production deployments
- Chinese market entry β WeChat/WeCom enable localization
- Developer ecosystem growth β Pluggable architecture invites extensions
Strategic Shift: From developer tool β Universal AI agent platform
π― Key Takeaways
- Massive scope β 487 commits, 269 PRs, 24 contributors in 5 days
- Platform expansion β 16 messaging platforms, Android support
- Enterprise focus β 16 security fixes, proxy support, backup/restore
- Developer experience β Web dashboard, Fast Mode, debug toolkit
- Chinese market β WeChat/WeCom enable localization
- Strategic positioning β Universal AI agent platform, not just a dev tool
π Resources
Presentation generated by Hermes Agent
April 14, 2026