Summary: Tacit Protocol is best understood not as just another AI-agent app or generic decentralized-identity toolkit, but as an attempt to define a distinct social layer for agents acting on behalf of humans. Its draft specification repeatedly separates six primitives that product pitches usually collapse together: Agent Cards for discoverable agent identity, encrypted Intent Broadcasting for what an agent is seeking, Authenticity Vectors for reputation/trust state, Match Scoring for compatibility evaluation, Intro Proposals for double-opt-in negotiation, and Progressive Context Reveal for staged disclosure after consent. That decomposition makes Tacit a useful comparison point for Agent Passport System, ENSIP-25, Next.ID, Orbis, Ethereum Follow Protocol, and other identity/social-graph systems because it exposes where control might actually sit in agent-mediated networking: DID choice, credential issuers, relay operators, authenticity-vector policy, match-ranking logic, or persona/reveal defaults.
What it does:
Defines agent-readable Agent Cards that publish an agent DID, supported Tacit protocol versions, transport endpoint, domain context, authenticity state, and optional preferences
Defines encrypted intent-broadcasting flows so agents can advertise what kinds of introductions, collaborators, clients, mentors, or services they are seeking without immediately revealing full human identity
Defines Authenticity Vectors as a multi-dimensional trust surface combining tenure, consistency, third-party attestations, and network-trust inputs rather than relying on self-reported bios alone
Defines match-scoring and intro-proposal flows where both represented humans must approve before a connection is revealed or completed
Enforces a three-layer identity model separating core human identity, persistent agent identity, and context-specific session personas
Uses DIDComm v2, W3C DIDs, and Verifiable Credentials as the reference transport and identity substrate while aiming to support multiple downstream applications such as B2B networking, local services, dating, and mentorship
Extends beyond discovery into agent-mediated commerce/service negotiation, making trusted introductions and later service interactions part of one broader protocol family
Key claims:
Tacit’s central claim is that MCP and A2A still leave a missing layer 3: agents can use tools and delegate tasks, but they still lack a standardized way to discover other agents, evaluate mutual trust, and negotiate introductions between the humans they represent.
The protocol’s main analytical value is its decomposition. Agent identity publication, intent discovery, trust computation, intro consent, and staged disclosure are treated as separate protocol surfaces, which is far more reusable for comparison than filing Tacit as a generic AI networking app.
The three-layer identity model is especially worth keeping. Tacit explicitly distinguishes core human identity, public network agent identity, and disposable session personas, which makes privacy policy and social-context disclosure legible as a separate control plane rather than as an app UX detail.
Authenticity Vectors are the project’s most distinctive proposed primitive. The whitepaper and spec describe them as cryptographically provable trust state derived from tenure, behavior consistency, attestations, and network-trust quality, which turns social trust into a structured protocol object rather than a loose profile score.
The double-opt-in and progressive-context-reveal flows matter because they define where the protocol stops and where human interaction begins. Tacit is not only a discovery directory; it is a consent and disclosure choreography for agent-mediated social matching.
The relay model is also important. Tacit says relay nodes should route encrypted messages and index discovery metadata without reading message contents, which makes relay policy and metadata visibility a distinct comparison layer beside DID and VC issuance.
Tacit belongs in the active corpus because it sharpens an emerging agent social trust branch that is easy to miss if everything agent-related is filed only under wallets, task delegation, or generic identity. It is effectively trying to make social discovery, compatibility, and intro consent into open infrastructure.
The strongest caveat is implementation maturity and governance specificity. The official materials are still draft-stage, concentrated in one repository and one author’s whitepaper/spec, and they leave important future control surfaces comparatively open — especially authenticity scoring policy, relay-network economics, and conformance governance.
Whitepaper: Yes. The canonical current whitepaper is docs/WHITEPAPER.md in the official repository, alongside the draft protocol specification collected in ../whitepapers/tacit-protocol-primary-sources-2026-05-15.md.