Used to confirm top-level positioning, linked documentation surfaces, and the project’s self-description as a trust layer for agent-mediated human introductions
Used to confirm the formal primitive split across Agent Cards, Intent Broadcasting, Authenticity Vectors, Match Scoring, Intro Proposals, Progressive Context Reveal, DIDComm transport, identity layers, and conformance language
Used as a lightweight confirmation that the project is presenting itself as something developers can integrate rather than only a narrative whitepaper
Key takeaways from primary sources
Tacit’s real contribution is not generic AI networking; it is a protocol proposal for turning social discovery, trust evaluation, and human-introduction consent into explicit machine-facing primitives.
The draft spec’s six-part decomposition is the main reason to keep the project in the corpus: Agent Cards, encrypted intents, authenticity vectors, match scoring, intro proposals, and progressive context reveal are separate control surfaces rather than one app-defined workflow.
The three-layer identity model is especially reusable. Tacit distinguishes core human identity, persistent network-visible agent identity, and context-scoped session personas, which makes disclosure policy a first-class layer instead of a UI afterthought.
The project’s most distinctive proposed primitive is the Authenticity Vector: a structured trust state combining tenure, consistency, attestations, and network-trust quality. This is analytically different from plain DID ownership, plain follower graphs, or plain verifiable-credential presentation.
Tacit uses DIDComm v2, W3C DIDs, and Verifiable Credentials as its default substrate, so it sits at an interesting junction between agent tooling, decentralized identity, and social-graph infrastructure.
The official materials repeatedly emphasize double opt-in and progressive reveal. That means the protocol is not just publishing profiles or credentials; it is trying to standardize how introductions are negotiated and how information becomes visible over time.
The commerce/service-negotiation positioning in the spec suggests Tacit wants to extend beyond introductions into trusted agent-mediated service exchange, which could make it relevant to later machine-commerce and agent reputation comparisons.
The biggest caveat is maturity and governance concentration. The sources are still draft-stage and mostly authored from one project repo, leaving important questions comparatively open around scoring governance, relay incentives, anti-spam policy, and conformance enforcement.
Specific source signals worth retaining
The README is the clearest compact source for the Layer 3 framing above MCP and A2A plus the quick-start developer positioning.
The protocol spec is the best source for the actual protocol objects and for the explicit requirement language around encryption, discovery, identity separation, and message flow.
The whitepaper is the best source for the broader use-case framing and for why the project thinks existing identity/social platforms are structurally insufficient.
The SDK package link matters because it shows the team is trying to surface Tacit as buildable protocol infrastructure rather than a purely conceptual research paper.
Why this entry belongs in the corpus
It sharpens an emerging agent social trust branch that would otherwise get flattened into generic AI-agent tooling or generic decentralized identity.
It gives the corpus a concrete way to compare social discovery, trust computation, consent choreography, and identity disclosure policy as separate layers.
The durable insight is that agent-mediated networking may become its own protocol surface: not just who are you?, but what are you seeking?, how is trust computed?, who decides when context is revealed?, and which relays or scorers become the actual chokepoints?