Self-hosted multi-model agent with JitRL memory, auditable council runs, SSH operations, live server screen, and macOS/Windows desktop client
agentd is a server agent that gets better as it works. No training, no GPU.
1. Put LLMs to work on your servers. Point it at any model — an OpenAI-compatible endpoint (your own relay counts) or native Anthropic / Gemini. Models act through real tools: SSH, shell, files. Dangerous commands pass a safety gate and wait for your click. 2. Watch it work. A real interactive terminal, a live view of the machine's screen, and a step-by-step trace, all in one browser console. One SSH tunnel reaches it; the server exposes no public ports. 3. Let it improve. It records what each action earned and does better next time in similar situations. You can verify that mechanism yourself with no API key at all.
- Want to run it now → 60 seconds - Want the numbers → Measured results - Want the mechanism → How it works
The Tauri desktop client and browser console now live in this same repository. One project, one API, one auditable trajectory.
Install it, then two commands to watch it learn. No API key, no network.
How to read it: without memory, the control arm wins 1 of 40 episodes. With memory, 11 — and the second-half success rate climbs. That is the memory working. For results beyond a single seed, see Measured results.
------ Learns while running (state, action, discountedreturn) triplets in SQLite; n-gram retrieval with an inverted index; advantage re-ranking; verbal self-critique stored and recalled. Per-memory credit accounting shows which memories actually get adopted Any model you want one baseurl + apikey + model — an OpenAI-compatible endpoint or a native Anthropic / Gemini se
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/hue913--agentd.html)
Shared, 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
nettanvirdev/inertiainertia-ai: a desktop workspace for AI agents. Chat, built-in tools, MCP and OpenAPI integrations, Docker/Daytona sandboxes with a live desktop, routines and memory. Bring your own model and keys.
polymind-inc/agent-framework-jsA TypeScript implementation of the Microsoft Agent Framework programming model, with semantic parity to the official .NET, Python, and Go implementations. Independent community project.
azimov777/casefileThe task tracker your AI agents keep for each other: every task carries a case file, so the next agent picks up where the last one stopped. MCP server + live board. Self-hosted, one-line install.
SysikNagibator/AiPCGive the agent "this" and he gets a real PC.
dhpradeep/aegisTurn your Claude subscription into a multi-tenant agent API with rate limits, cost tracking, autonomous objectives, and an admin dashboard. Self-hostable, no API key.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.