CapToken

Capability tokens for governed agent and tool authorization.

CapToken defines lightweight capability tokens for authorizing what agents and tools are allowed to do, under explicit scope, limits, delegation rules, and expiry constraints.

As agent systems become more modular, permission boundaries often remain implicit. Agents gain access to tools, resources, and actions without a portable, inspectable authorization layer.

CapToken exists to make those boundaries explicit.

Why CapToken

Many agent systems still rely on weak or implicit authorization assumptions:

• broad tool access
• missing scope boundaries
• unclear delegation rights
• weak expiry semantics
• no portable capability format
• unclear limits around cost, duration, or request volume

CapToken introduces a lightweight capability token layer for bounded agent execution.

Core goals

• define portable capability tokens
• validate token structure
• clarify authorization scope
• encode limits and expiry
• represent delegation permissions explicitly
• support inspection and dev-friendly tooling

Initial scope

CapToken v0 starts with:

• captoken.yaml parsing
• token validation
• human-readable explanation
• CLI commands
• example tokens

Later phases may add:

• signing
• revocation
• token bundles
• policy packs
• runtime adapters
• coordination with contracts, traces, and skill reliability layers

Example

captoken validate ./examples/web-research-token
captoken inspect ./examples/web-research-token
captoken explain ./examples/web-research-token
captoken mint-dev ./examples/web-research-token

Philosophy

CapToken is not a full IAM platform.

It is a portable specification and validation layer for bounded agent and tool authorization.

Roadmap

See:

• docs/vision.md
• docs/architecture.md
• docs/token-spec.md
• docs/roadmap.md