Your agent is acting on your behalf. Make it count.

Sign up at dictiva.com/sign-up. Provision an API key — the API Keys guide walks you through the dashboard path, or use the fast path documented in the MCP server guide.

Drop this .mcp.json at your project root:

{
  "mcpServers": {
    "dictiva-governance": {
      "type": "streamable-http",
      "url": "https://dictiva.com/api/mcp",
      "headers": { "Authorization": "Bearer dv_live_..." }
    }
  }
}

Verify the connection:

claude mcp list
# dictiva-governance ... ✓ Connected

From inside Claude Code, ask:

Use the dictiva-governance MCP to list applicable policies for an agent.

Claude calls list_applicable_policies_for_actor with actor_type: "agent". The response is a tight list — only statements scoped to agents, only the enforcement modes you can actually act on:

{
  "data": [
    {
      "displayId": "AAG-007",
      "title": "Agents must declare adopted standards",
      "enforcementMode": "machine_readable",
      "applies_to": "agent",
      "actor_scope": "agent-workflow"
    },
    { "displayId": "AAG-014", "title": "Code-reviewer agents commit to review SLA", "...": "..." }
  ]
}

Pick one statement to commit to. Note its displayId — you'll reference it in the next step.

Pick one of three patterns, depending on where the work lives. The artifact's content becomes the evidence — its existence in your repo is the claim that you're committing to the named statement at the declared tier.

Pattern A — A skill (.claude/skills/<name>/SKILL.md):

---
name: my-skill
description: ...
attests: AAG-007 tier=T1
---

Pattern B — A memory file (memory/<name>.md or .claude/memory/<name>.md):

---
name: my_memory_entry
description: ...
type: feedback
attests: AAG-007 tier=T1
---

Pattern C — An AGENTS.md section (anywhere in your repo). Since AGENTS.md has no file-level frontmatter, the convention is an HTML-comment marker immediately above a section heading:

<!-- attests: AAG-007 tier=T1 -->
## My agent's working conventions

This agent operates under...

The grammar is the same in all three: <statementRef> tier=T<n>. Multiple statements stack — use a YAML list (attests: [AAG-007 tier=T1, AAG-014 tier=T1]) or repeat the HTML marker. The full grammar is in the authoring guide.

Two paths. Pick the one that fits your workflow.

Path A — Live, from Claude Code. Ask Claude to call attest_statement:

Use dictiva-governance to attest_statement for AAG-007 at tier T1, with my skill at .claude/skills/my-skill/SKILL.md as evidence. My agent DID is did:web:example.com:agents:agent-my-claude.

The MCP tool issues a signed W3C Verifiable Credential and returns the attestation id immediately. This path needs a Business plan or higher (MCP-tool access) and a registered agent identity in your tenant.

Path B — Via commit trailer. Add this line to the trailing block of your commit message:

Attests: AAG-007 tier=T1 evidence=.claude/skills/my-skill/SKILL.md

Canonical grammar (matches the parser at webapp/src/lib/attestation/attests-trailer.ts):

^Attests:\s*([\w-]+)\s+tier=(T[1-6])(?:\s+evidence=(\S+))?\s*$

The trailer must be on its own line in the commit body's trailing block — the same convention as Co-Authored-By: or Signed-off-by:. When the merged-PR webhook lands (currently founder-gated, shipping shortly), each Attests: line creates a draft credential in your tenant's drafts inbox; you finalize with one click.

A note on tiers: T1 is plausible from this workflow today. T2 requires a semantic_similarity_report artifact that links your prose to the statement — that tooling hasn't shipped yet. T3 (Referenced) and T4 (Procedural) are reachable when your skill, memory file, or ADR demonstrably codifies the commitment. T5 and T6 require scope expressions and runtime enforcement; they typically come from infrastructure, not solo authoring.

From Claude Code:

List my attestations from the dictiva-governance MCP, filtered to my agent.

Calls list_my_attestations. The response shows your live credentials, their tiers, the evidence you submitted, expiry, and revocation status:

{
  "data": [
    {
      "id": "uuid-...",
      "subjectDid": "did:web:example.com:agents:agent-my-claude",
      "statementDisplayId": "AAG-007",
      "commitmentTier": "T1",
      "issuedAt": "2026-04-27T...",
      "expiresAt": "2027-04-27T...",
      "evidence": [{ "evidenceKind": "skill", "reference": ".claude/skills/my-skill/SKILL.md" }]
    }
  ]
}

Or open the attestation page on your tenant: Workbench → Attestations.