tf-types
Core semantic types, traits, and schemas powering the TrustForge protocol.
Part of TrustForge — the open-source trust fabric for
AI-native software, devices, and verifiable action. The Rust crates and
the TypeScript packages (@trustforge-protocol/*) are mirrored
reference implementations kept in lockstep by a cross-language
conformance suite.
Install
Overview
TrustForge type bindings and semantic core. Generated wire types live
under generated/; hand-written semantic helpers live as sibling
modules:
- Identity & authority — actor / instance IDs, trust domains, capabilities and negative capabilities, delegation chains, revocation, quorum, approval ceremonies.
- Wire formats —
.tfpkt/.tfbundlebinary containers, packets, envelopes, proof events, session primitives. - In-house codec layer (mirrored 1:1 with the TS package, gated by
cross-language conformance vectors): canonical JSON (
canonical), deterministic CBOR (cbor), TF-YAML strict subset (yaml), compact JWS/JWT (jws), base64 (encoding), and the capability glob language (glob). - Bridges — WebAuthn, SPIFFE, OAuth/GNAP, MCP/A2A, TLS, DID, Matrix, service-mesh compatibility projections.
- Policy — the native policy engine, agent guard, and manifest
loaders for
.tf/(agent contract, threat model, policy).
Cryptographic primitives are never implemented here — signature, AEAD,
hash, and KDF math delegates to reviewed crates (ed25519-dalek,
p256/p384, rsa, chacha20poly1305, sha2, blake3, argon2).
Links
- API docs: docs.rs/tf-types
- Source: crates/tf-types
- Specs & conformance vectors: KodyDennon/TrustForge
- Issues: KodyDennon/TrustForge/issues
Status
Draft — experimental. Apache-2.0.