Pulse — Hermes-style self-improving AI agent. Reliability-first rebuild with evaluated skill self-evolution, multi-agent team orchestration, dialectic user modeling, and fully self-hosted default stac
Pulse — Self-improving AI Agent (Reliability-First)
A self-improving personal AI agent with a reliability-first core. Compatible with the agentskills.io open standard. Fully self-hostable by default — Ollama + SQLite FTS5, zero cloud dependency.
Feature Detail ----------------- Web UI Session management dashboard on port 10000 — browser-based chat, tool listing, settings Social Gateways Feishu, WeChat, WhatsApp — async webhook-based bridges to external chat platforms 10 Built-in Tools websearch, webfetch, writefile, editfile, pythonexec, shellexec, httpclient + original read/list/dir/calc Dynamic Tools Drop .yaml, .json, .py into ~/.pulse/tools/ — auto-registered at startup Executable Skills Skills can declare tools, self-test, hot-reload — BaseExecutableSkill + SkillHandle Session Memory Multi-turn conversations within same session; previous turns auto-injected as context Streaming Output Orchestrator.runstream() yields token chunks; TUI shows live spinner Plan→Execute→Verify System prompt drives step-by-step planning with DANGEROUSTOOLS audit Feedback Loop addcorrection("always use type hints") → remembered in future system prompts Reliability-first Every LLM/tool call wrapped in classified error recovery + exponential backoff + hard token budget guardrail Provider Fallback Ollama → OpenRouter → Anthropic chain + token-bucket rate limiter Skill Evaluation Golden-task replay before promote/quarantine/rollback; only complex multi-tool tasks trigger evolution File Logging ~/.pulse/logs/pulse.log with daily rotation (7 days retained) Container Ready Dockerfile + .dockerigno
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/Alex663028--pulse-agent.html)
Self-hosted, local-first AI assistant server: swappable LLM providers, review-gated memory, an encrypted vault, and wake-word voice. No accounts, no telemetry.
693696817/mcd-roundtable麦门圆桌（mcd-roundtable）｜五个 AI 人格辩论，逼出最优麦当劳订单。基于麦当劳 MCP（mcd-mcp / Model Context Protocol）的多智能体点餐决策 Skill：菜单、券、营养、价格全部来自官方接口，金额只认 calculate-price，绝不估算。凑单最省 / 优惠券最大化 / 营养配餐 / 一键下单。Multi-agent McDonald's MCP
leluong141996-dev/Agent-Brain-HubShared, brain-inspired memory for AI agents: 13 brain regions with awake/sleep loops, a live visualizer, agents connected via REST/SDK/MCP, governance & audit, and business-value metrics. Works with C
rokbenko/quackd🦆🧠 Give your small robot a brain. Tell a Microduck, a Reachy Mini, a LeRobot arm or any ROS base what you want in plain language. An LLM (Claude, OpenAI, Gemini, Grok, or a local model via Ollama, vLL
bybit-exchange/kaasTurn scattered notes, docs and transcripts into a queryable Markdown wiki — an LLM knowledge-base compiler with MCP access, no embeddings, self-hosted.
cmoraes10/lociTyped long-term memory for AI agents. One core, an MCP server, and a Hermes plugin.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.