TOOL DISCOVERY
Veritan exposes a governed set of MCP tools that AI agents can discover, inspect, and invoke. No undocumented surfaces.
DEVELOPERS · MCP
AI agents connect
through MCP.
AI agents connect to governed Veritan capabilities through the Model Context Protocol. Tool discovery, scoped authorization, and structured evidence-bearing responses — designed for machine consumption.
MCP IS FIRST CLASS
MCP is not an optional developer appendix. It is a primary access surface for AI agents that need to know what is true.
WHY MCP
AI agents need governed, evidence-bearing tool responses.
When an AI agent queries Veritan through MCP, it does not receive an assertion — it receives a structured, evidence-bearing response it can evaluate. The tool call itself is authorized, scoped, and audited. The response includes provenance, authority, and state alongside the data.
AUTHORIZATION
Every tool invocation is authorized against the credential scope. Agents cannot exceed their granted capability set.
EVIDENCE RESPONSES
Every tool response includes machine-readable evidence so the agent — and anything it reports to — can verify the result.
TOOL MODEL
Governed tools. Structured responses.
Veritan MCP tools follow a consistent naming and response convention. Tool names are stable within a version. Responses include the requested data and evidence fields in a machine-readable schema the consuming agent can inspect.
EXAMPLE MCP TOOL — ILLUSTRATIVE FORMATMCP
# Tool discovery
tools/list → [
{ name: "veritan.records.verify", description: "..." },
{ name: "veritan.provenance.inspect", description: "..." },
{ name: "veritan.authority.check", description: "..." }
]
# Tool invocation
tools/call veritan.records.verify
{
"record_id": "<record-id>",
"evidence_scope": ["provenance", "authority", "state"]
}
→ {
"record": { "id": "...", "state": "active" },
"evidence": { "provenance": {...}, "authority": {...} },
"verified_at": "2026-08-18T00:00:00Z"
}Tool names follow the format veritan.<domain>.<action>. Actual tool names are confirmed at integration time against the canonical MCP registry.
INTEGRATION FLOW
From agent to evidence.
An AI agent connecting to Veritan through MCP follows a governed flow: credential authentication, tool discovery, scoped invocation, and evidence-bearing structured response.
1. AUTHENTICATE
The agent presents a project credential with defined scope. Authentication is validated at the MCP server layer before tool discovery is permitted.
2. DISCOVER
The agent calls tools/list to discover available Veritan tools. Only tools within the credential scope are returned.
3. INVOKE
The agent calls a specific tool with a structured input. The invocation is authorized against scope, logged, and executed.
4. RECEIVE EVIDENCE
The response includes the structured result plus machine-readable evidence: provenance chain, authority record, state, and verification timestamp.
5. AUDIT
Every invocation is recorded in the platform audit log. The agent, operator, and downstream systems can inspect what was called and when.
AUDITABILITY
Every agent interaction is audited.
Veritan MCP is not a pass-through proxy. Every tool invocation is a first-class platform event: logged, scoped, signed, and retained. Operators can inspect the complete record of agent interactions via the Console or the API.
INVOCATION LOG
Every tool call: timestamp, credential, tool name, input scope, response hash, and outcome.
SCOPE ENFORCEMENT
Credentials cannot invoke tools outside their granted scope. Attempted overreach is logged and blocked.
EVIDENCE CHAIN
The evidence returned by each tool call is retained and linked to the invocation record for downstream verification.