Agent Passport System skill
Open protocol giving agents signed identity, delegation that only narrows, and receipts for every allowed or denied action.
- 64 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Commerce and spending modules involve money; limits and keys must be configured by the user.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/agent-passport-systemDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Commerce and spending modules involve money; limits and keys must be configured by the user.
This is third-party code. Review the repository files before installing.
What it does
Agent Passport System is an open protocol with a TypeScript reference implementation that gives an agent limited authority and records what it did with it. An Ed25519 passport says who the agent is. A delegation chain says who authorized it, and passing rights on can only narrow them. A gateway checks scope, spending limits, expiry and approved services before anything runs. Every decision, allowed or denied, gets a signed receipt that anyone can verify later. The repository ships agent skills and the `agent-passport-system-mcp` MCP server: 154 tools by default, 28 with `APS_PROFILE=essential`.
Who it is for. Teams that hand agents access to payments, data or external services and want a verifiable trail of actions.
Good fit when
- An agent may issue refunds up to a fixed amount and no more.
- One agent hands a task to another and rights must not grow along the way.
- An auditor needs to see which actions were allowed, which denied and why.
- Access is revoked for a parent agent and all its descendants at once.
Not a fit when
- The agent works alone on local files and there is nothing to delegate.
- You need a finished product with a UI: this is a protocol and libraries.
- Cryptographic receipts are overkill for a one-off script.
Example request
Create a passport for a helper agent, delegate refunds up to 100 dollars and show a 200 request being denied with a receipt.Limitations
This is a protocol and libraries, not a ready control system: the gateway, keys and receipt storage are set up and kept by the user. Python, Go and Rust versions live in separate repositories and are checked against the reference with conformance tests. The MCP `full` profile exposes 154 tools, which is heavy for context, so start with `essential`. Commerce and spending modules check limits, but real payments go through your own connected services.
How to disable. Remove the skill and the `agent-passport-system-mcp` MCP entry from the client. Passport keys and issued delegations are kept and revoked separately.
FAQ
What is a receipt?
A signed record of a gateway decision: what was requested, which authority was presented, which policy applied and what was decided. It can be verified locally in the browser on the project's verification page.
How do MCP profiles differ?
`essential` shows 28 tools: identity, delegation, enforcement, commerce and reputation. `full` includes all 154. Other profiles such as `governance` or `data` assemble the groups you need.
Which other languages are supported?
Python, Go and Rust, each in its own repository. A conformance suite checks them against the TypeScript reference.
Is a GitHub token required?
No. The `GITHUB_TOKEN` variable in the skill is marked optional and is only used by one tool that publishes to the public agent registry.
Where to start after installing?
Import from `agent-passport-system/core`: it has 24 functions such as `createPassport`, `createDelegation` and `evaluateIntent`, enough for most integrations.
Discussion
No questions or reviews yet.
Related
SkillsA utility keeps keys in the system keychain and injects them at the transport layer, so the agent never sees a value.
An MCP server with 47 tools: signed AI agent identities, W3C credentials, delegated permissions, an action log and EU AI Act documents
AgentSafe is an agent API gateway that links an executed operation to a single-use grant and a record in its audit trail.
An MCP memory server for agents plus a free gate that stops dangerous commands before they run
An open OWASP standard where a separate guardian agent inspects an AI agent's action before it runs and allows, denies or edits it
A policy gateway for AI agents: it checks tool calls before they run, holds risky ones for approval and writes an audit trail
README in short
The protocol core: passports, delegation, a gateway and signed receipts. On top there are governance, commerce and agent-to-agent modules plus adapters for popular frameworks.