Connect AI agents
to Synaura via MCP.

Quick Start

Point any MCP-compatible client at your organization's endpoint with a bearer token. No SDK installation required.

Endpoint

MCP Server URL
https://mcp.synaura.ai/YOUR_ORG_ID

Authorization
Bearer YOUR_TOKEN

Claude Desktop / Claude Code

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "synaura": {
      "url": "https://mcp.synaura.ai/YOUR_ORG_ID",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Cursor

In Cursor settings, add an MCP server with the same URL and authorization header. Cursor supports the MCP standard natively.

Any MCP runtime

Any client implementing the Model Context Protocol specification can connect. The endpoint supports tools/list, resources/list, resources/read, and tools/call.

Tools Available

Calling tools/list returns procedure-based tools dynamically generated from your organization's knowledge base. Each tool includes an _alignment metadata object (documented below) so agents can make trust-aware decisions before executing.

synaura.ask

The primary tool for graph-grounded Q&A. Ask any question about your organization's processes, decisions, or procedures, and receive an answer grounded in the knowledge graph — not hallucinated from training data.

// Example tool call
{
  "name": "synaura.ask",
  "arguments": {
    "question": "What is our deployment procedure for production releases?"
  }
}

// Response includes source references and alignment metadata
{
  "answer": "Production releases follow the three-stage deploy...",
  "sources": ["proc_deploy_prod", "dec_v4_03"],
  "_alignment": { "drift_score_7d": 0.05, "trust_status": "current", "latent_risk_flag": false }
}

Procedure-based tools

Each published procedure in your organization is exposed as an individual tool. Agents can discover what your organization knows how to do, and invoke procedures with full context.

Resources Available

MCP resources provide read access to structured data in your Synaura organization. Use resources/list to discover available resources, or resources/read with a specific URI.

Alignment Metadata

Every tool returned by Synaura's MCP endpoint includes an _alignment object. This metadata lets agents make informed decisions about whether to trust and execute a tool, or to pause and ask for human review.

What each field means

• drift_score_7d — How much observed signals have diverged from the documented procedure over the last 7 days. 0.0 means perfect alignment; values above 0.3 indicate meaningful drift.
• trust_status — A categorical assessment: "current" means the procedure is up-to-date and actively followed; "stale" means it hasn't been reviewed recently; "review_needed" means drift or signals suggest the procedure may no longer reflect reality.
• latent_risk_flag — Boolean. When true, Synaura has detected a latent risk: the procedure exists and looks correct, but real-world signals suggest it may not be followed or may produce unexpected outcomes.

Sample agent policy

Use alignment metadata to build guardrails into your agent's behavior:

// Before executing any Synaura tool, check alignment
if (tool._alignment.trust_status !== "current") {
  require_human_approval();
}

if (tool._alignment.drift_score_7d > 0.3) {
  warn("Marketing claims may differ from reality");
}

if (tool._alignment.latent_risk_flag) {
  escalate("Latent risk detected — procedure may not reflect actual practice");
}

Provenance

Every artifact returned through MCP — procedures, decisions, process cards — is cryptographically signed using ML-DSA-65 (FIPS 204), the NIST-standardized post-quantum signature algorithm. Agents can verify artifact integrity independently.

Verification URLs

Each signed artifact has a public verification URL that returns the signature, public key, and content — enough for independent verification without Synaura access:

// Public verification endpoints (no authentication required)
GET https://protocol.synaura.ai/verify/process_card/{id}
GET https://protocol.synaura.ai/verify/decision/{id}
GET https://protocol.synaura.ai/verify/procedure/{id}

Verification is performed client-side using the @noble/post-quantum library. The signature is checked against the public key and the canonical payload. No trust in Synaura's servers is required — the artifact is self-proving.

What this means for agents

• Agents can verify that the knowledge they received has not been tampered with since it was signed.
• The signer identity (signer_user_id) is cryptographically bound — agents know exactly who authored or approved the content.
• ML-DSA-65 is quantum-resistant, so signatures remain valid through the post-quantum transition.