What is this?
Personal Agent Proxy is an open-source, self-hostable access layer that lets you connect your personal services once, then give each AI agent only the specific permissions it needs, while auditing everything that agent does.
Personal Agent Proxy is an open-source, self-hostable access layer that lets you connect your personal services once, then give each AI agent only the specific permissions it needs, while auditing everything that agent does.Is this a hosted service or something I run myself?It's a self-hostable proxy you deploy and run on your own infrastructure, sitting between your connected services and whichever AI agents you give access to.
It's a self-hostable proxy you deploy and run on your own infrastructure, sitting between your connected services and whichever AI agents you give access to.How do I connect my services and set per-agent permissions?You self-host the proxy, connect each of your services to it once, and then configure the specific permissions each AI agent is allowed to use against those services.
You self-host the proxy, connect each of your services to it once, and then configure the specific permissions each AI agent is allowed to use against those services.Which AI agents does it work with?It is designed to work across agents including Dots, Grok Bot, Muse, Instinct, Poke, Claude, ChatGPT, and Hermes Agent.
It is designed to work across agents including Dots, Grok Bot, Muse, Instinct, Poke, Claude, ChatGPT, and Hermes Agent.Does it stop an agent from quietly getting more access over time?Yes, that is the core design goal: each agent only receives the granular permissions you explicitly grant, so a provider cannot get an agent's full private context or silently expand what it is allowed to do.
Yes, that is the core design goal: each agent only receives the granular permissions you explicitly grant, so a provider cannot get an agent's full private context or silently expand what it is allowed to do.What does 'auditing activity' mean in practice?The proxy keeps a record of what each connected AI agent does through it, so you can review agent actions after the fact rather than just trusting the agent.
The proxy keeps a record of what each connected AI agent does through it, so you can review agent actions after the fact rather than just trusting the agent.Why not just give each agent direct access to my accounts?Direct access usually means handing over your full private context to every provider; this proxy lets you scope each agent down to only the permissions it actually needs.
Direct access usually means handing over your full private context to every provider; this proxy lets you scope each agent down to only the permissions it actually needs.What are similar projects?CanvasTTY, evenfire, anythingmcp, mimiclaw, OpenAPPA, brigade, casbin-gateway, botdirectory.ai
CanvasTTY, evenfire, anythingmcp, mimiclaw, OpenAPPA, brigade, casbin-gateway, botdirectory.aiWhich topics is it in?muse, permission, personal-agent, access-control, authorization, audit-log, grok, privacy-first, chatgpt, proxy, agents, claude
muse, permission, personal-agent, access-control, authorization, audit-log, grok, privacy-first, chatgpt, proxy, agents, claude