@receiz/mcp-server exposes the canonical Receiz SDK rails as agent-callable tools. Agents can verify proof objects, resolve public app-state, bootstrap tenant sessions, publish delegated additions, inspect capabilities, and build production Receiz apps without inventing a parallel integration layer.
Receiz v129.0.0 exposes one coordinated proof-native runtime: source, identity, ownership, lawful action, and verified history remain bound to the object. Capability discovery covers 704 callable SDK exports, 425 client methods, 356 explicit MCP tools, and 43 AI skills. Existing v123, v124, and v125 tool names retain their exact protocol identities beneath sealed proof authority. Read the current release outcomes.
The object is authority. Receiz proof objects, sealed artifacts, and verified local truth remain the source of truth. Public proof rails and SDK/API appends expose or extend that truth beneath the proof object. MCP exposes those rails to agents and must never replace, rename, or outrank identity, ownership, settlement, public proof, or verification primitives.
The edge verifies and retains the object. Remote services perform scoped deterministic mutation, while database storage provides projection, sync and recovery beneath the admitted proof head.
Install
Install the MCP server next to the SDK. The server wraps @receiz/sdk; it does not carry a separate data model.
npm install @receiz/mcp-server @receiz/sdk
Run
Public reads can run without a token. Delegated writes require a Receiz Connect/OIDC access token with the scopes for the rail being called. Create the MCP token from Developers Connect using Create Delegated Agent Token.
Run receiz_doctor and receiz_capabilities to inspect the app, tenant host, callback URLs, scopes, and available rails.
Use receiz_required_scopes to request the exact OIDC scopes needed for identity, wallet, commerce, app-state, proof, media, domains, customers, and merchants.
Use receiz_mcp_login to open the delegated agent-token path, then use receiz_authorize_url or receiz_ensure_tenant_session when user consent or tenant bootstrap is required.
Use receiz_public_store_resolve or receiz_app_state_resolve for cold-start first paint from public projection truth.
Publish verified additions with receiz_public_store_publish or receiz_app_state_publish using delegated access.
Use receiz_webhook_events_catalog, receiz_webhook_register_endpoint, and receiz_webhook_receiver_scaffold for signed event delivery.
Use receiz_artifact_verify before receiz_artifact_extract_verified, and use receiz_artifact_round_trip_check to prove exact complete-byte custody. Every artifact verdict comes from the shared SDK verifier.
For embedded identity, call receiz_v123_identity_exchange_proof_authority only after edge verification and explicit in-application consent; it does not require redirecting the person to receiz.com.
Plan canonical public world operations with the v123 MCP planner tools, execute Phi on exactly one Settlement or Reserve rail, and resolve committed, zero-write, or unknown before retrying an ambiguous result.
MCP delegated tokens are for agent and server operations. Merchant-owned storefront state should use the merchant Identity Seal or Receiz Key through the SDK signed publish rail. That keeps public-store sync beneath the merchant proof object and avoids treating an agent token as merchant authority.
Use MCP to diagnose, inspect, and resolve. Use publicStore.publishWithIdentityProof when the merchant proof object is the publish authority.
Agent hosts can expose stable Receiz URI templates for public proof assets, pitch witnesses, public store projections, SDK docs, and proof object schema guidance.