Security sidecar between an AI coding agent and everything it touches — MCP servers, LLM APIs, SSH, subprocess env. Release mirror; development is private.
Universal security sidecar for AI agents.
agent-sidecar sits between your AI coding agent and everything it touches — MCP tool servers, LLM APIs, SSH hosts, databases — enforcing security policy, brokering short-lived credentials, scanning for prompt injection, and recording every action in a keyed, hash-chained audit trail.
It proxies real MCP servers: one persistent process per upstream with a proper initialize handshake, real tool schemas passed through, and policy applied to every call. LLM tool calls are enforced for Anthropic and OpenAI in both streamed and buffered responses — neither stream: false nor n: 2 is a way around policy. A stream whose shape it does not recognise is forwarded and says so, in the log and in the audit trail, rather than leaving you quietly unprotected.
Where a surface is not covered in a given session, the wrapper says that too, inline and in the closing summary. A run that guarded nothing must not look like a run that guarded everything.
Surface How Status --------- MCP tools Persistent session per upstream server, real handshake, real inputSchema. Tools re-advertised as .; policy, secrets and audit on every call Working — run reports it plainly when it finds no MCP config to take over Every MCP scope Claude Code is launched with --strict-mcp-config, so a wrapped session reads only the config agent-sidecar generates — user, local and project scopes are all covered, and no config file is modified at all Working — the scope list is measured against the real binary by scripts/probereality.sh, not read off documentation. Agents without the f
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/agent-sidecar--agent-sidecar.html)
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is
zhuyansen/new-term-radarDaily radar for new AI terms: names of new models, products and repos, caught on X, Hugging Face, GitHub and sitemaps before keyword tools list them.
yzg1226/mcd-party-officer🍟 麦当劳门店派对成团调度 Skill — 把派对从「策划」推进到「办得成」：成团概率雷达 / 包场vs拼团 / 择日黄历 / 儿童营养指数 / 招募卡与邀请函
fellowgeek/mcp-memoryAn OKF-backed Model Context Protocol (MCP) server delivering persistent long-term memory and SQLite FTS5 search for AI agents.
hirotomasato/yowesGenerate realistic teacher documents (ID cards, licenses, letters) for 13 countries via MCP.
hujjun/mcd-saver-strategist麦当劳省钱点餐助手（麦门精算师）｜麦当劳优惠券组合优化 Skill：拉取麦当劳菜单与全部可用优惠券，算出最省的点餐组合，巨无霸套餐 ¥47 → ¥21.9（省 53%）。McDonalds MCP（麦当劳 MCP）coupon optimizer · McDonalds China MCP Server · Agent Skill for WorkBuddy · 麦当劳程序员创意开发大赛参赛作品
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.