Turn 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.
Turn your Claude subscription into a multi-tenant agent API — with a dashboard.
Aegis is a self-hostable server that exposes the Claude Agent SDK over your Claude Max/subscription login (via the bundled claude CLI) instead of an ANTHROPICAPIKEY. It adds per-tenant API keys, rate limiting, cost tracking, isolated session workspaces, autonomous objective loops, MCP servers, an OpenAI-compatible chat endpoint, and a full admin console on top of the SDK — so several tenants can safely share one subscription-backed agent runtime.
[!IMPORTANT] Auth model — read before deploying. Aegis drives Claude by shelling out to a claude CLI logged into a personal/Team subscription, not by billing an Anthropic API key. Anthropic's terms discourage reselling or provisioning subscription access to third parties. This project is intended for personal, internal, or trusted-team use of your own subscription. To run it as a commercial multi-tenant service, switch billing to API keys — that's a config change in the runtime layer, not a rearchitecture.
- Runs on your subscription — no API key; the Claude CLI is bundled with the SDK, so there's nothing extra to install. - Multi-tenant — tenants, scoped API keys, per-key rate limits (RPM) and daily cost caps. - Agents — configurable model, reasoning effort, allowed tools, MCP servers, permission mode, and autonomous bypass. - Sessions — stateful chats with isolated workspaces, file upload/download, and a live streaming chat UI. - Objectives — autonomous goal-driven loops graded by a separate LLM evaluator until success or budget. - MCP servers — re
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/dhpradeep--aegis.html)
Self-hosted documentation RAG pipeline that serves semantic search to LLM agents over the Model Context Protocol.
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.
rokbenko/quackd🦆🧠 Give your Microduck a brain. Tell a small robot with two legs what you want in plain language. An LLM (Claude, OpenAI, Gemini, Grok) uses the skills it already has to do it. Includes a simulator, .
chris-asmussen/wardenOne MCP server that fronts many MCP servers + Skills behind a small tool set and routes to the best one, keeping an agent's context tiny. Migrate your Claude Code MCPs/skills behind it.
KernelLord/pickysteveSkill router and context picker for Claude Code, Cursor, and Aider. Auto-picks the right skill for every prompt, with prompt-injection filtering for MCP.
MuhittinYilmazer/akanaSelf-hosted, local-first AI assistant server: swappable LLM providers, review-gated memory, an encrypted vault, and wake-word voice. No accounts, no telemetry.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.