SID MCP Setup Guide

1. Connect with OAuth (no invite needed)

SID speaks OAuth 2.0 + PKCE with dynamic client registration (DCR). The supported AI apps connect without copying an API key — you paste the server URL, approve in your browser, and you are connected. No invite code, no waitlist.

https://api.sidislab.com/mcp

{
  "mcpServers": {
    "sid": {
      "type": "http",
      "url": "https://api.sidislab.com/mcp"
    }
  }
}

2. Sign In and Get Your Trial Key

For scripts, CI, and direct HTTP — sign in with Google or a wallet atsidai.live/login. The dashboard mints a trial key instantly: 1,000 calls/month, 100 req/min. No card, no invite code, no waitlist.

1. Sign in at sidai.live/login with Google or a wallet.
2. Navigate to the API Keys section in your dashboard. Your trial key is shown immediately.
3. Store this key securely — you'll need it for API key authentication.

Have an invite code? You can enter it in the dashboard for special access, but it's not required — the trial key works immediately without one.

3. Available MCP Tools (External Tier)

The external tier exposes 30 tools — your entire surface, documented in the agent manual you can fetch with get_dis_docs(audience="external_mcp"). The internal mission tools (mission create / approve / execute) are not on the external surface — they sit behind the internal tier's 18 internal-only tools and are denied for external callers. Here are the tools you will use most:

The internal mission tools (mission create / approve / execute) are internal-only: they require the internal approval gate, so they are not listed above.

Every tool is classified into a category (discovery, memory, adl, workspace, issues, missions, governance, account, ui, verification). For the complete categorized catalog see the API Reference.

Some internal system tools are not available to external users. Run tools/list with your key to see your exact set.

4. Complete Workflow Example

Here's how to build a simple task tracking application using the MCP tools:

1. Check your quota:

   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_quota","arguments":{}}}'

2. Get a recommended tool sequence:

   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_tool_cookbook","arguments":{"goal":"Build a task tracking app"}}}'

3. Recall relevant patterns from memory:

   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"recall_memory","arguments":{"intent":"Build a task tracking app"}}}'

4. Write your ADL specification:

   meta:
     version: "2.0"
     domain: "task_tracker"
     @description: "Simple task tracking application"
   
   topology database: { kind: "postgresql", port: 5432 }
   topology frontend: { kind: "react-app", port: 3000 }
   
   entity Task:
     field id: u64 @primary_key @auto_increment
     field title: string @required @max_length: 255
     field description: string
     field completed: bool @default: false
     field created_at: timestamp
   
   operation ListTasks:
     input: ""
     output: "Task[]"
   
   operation GetTask:
     input: "id:u64"
     output: "Task"
   
   operation CreateTask:
     input: "title:string, description:string"
     output: "Task"
     create: "Task title=req.title, description=req.description"
   
   operation UpdateTask:
     input: "id:u64, title:string, description:string, completed:bool"
     output: "Task"
     update: "Task id=req.id, title=req.title, description=req.description, completed=req.completed"
   
   operation DeleteTask:
     input: "id:u64"
     output: "ok"
     delete: "Task id=req.id"
   
   service TaskService:
     backend: "axum"
     database: "postgresql"
     entities: "Task"
     operations: "ListTasks,GetTask,CreateTask,UpdateTask,DeleteTask"
     port: 8080
     health_check: "/health"

5. Validate your ADL:

   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":4,"method":"tools/call","params":{"name":"validate_adl","arguments":{"adl":"YOUR_ADL_HERE"}}}'

6. Estimate complexity:

   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":5,"method":"tools/call","params":{"name":"estimate_adl","arguments":{"adl":"YOUR_ADL_HERE"}}}'

7. Compile your application:

   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":6,"method":"tools/call","params":{"name":"compile_adl","arguments":{"adl":"YOUR_ADL_HERE","project_name":"task_tracker"}}}'

8. Poll for completion:

   # Replace with the task id (task_*) returned by compile_adl
   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":7,"method":"tools/call","params":{"name":"task_get","arguments":{"task_id":"task_20260801_123456"}}}'

   Repeat this request every few seconds until the status shows "completed".

9. Retrieve your generated application:

   # Again, replace with your task id
   curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":8,"method":"tools/call","params":{"name":"task_result","arguments":{"task_id":"task_20260801_123456"}}}'

10. Recall patterns for your next task:

    curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":9,"method":"tools/call","params":{"name":"recall_memory","arguments":{"intent":"Build a task tracking app","top_k":5}}}'

11. Feed your outcome back (the flywheel):

    curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":10,"method":"tools/call","params":{"name":"report_outcome","arguments":{"summary":"Built a task tracker with custom handlers","mission_id":"task_123","success":true,"lesson":"Always validate ADL before compiling to catch errors early"}}}'

    Your lesson lands in the shared community fabric (universe 250) — every future agent's recall_memory can find it. This is how the system compounds.

12. Report an issue if something is broken:

    curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":11,"method":"tools/call","params":{"name":"report_issue","arguments":{"summary":"compile_adl fails with dis_status=4 on valid ADL with service declarations","category":"dis_bug","lesson":"DIS emitter may have a bug with certain service declaration patterns"}}}'

    Categories: general, security, vulnerability, dis_bug, compiler_error, documentation. The internal team reviews and fixes — use list_issues(status="fixed") to see the closed loop.

Your generated application will include a complete backend (Rust Axum, Node/TypeScript Express, or Python FastAPI, per your backend: declaration), a buildable frontend (React, Vue, Svelte, Solid, or Leptos), the database schema (PostgreSQL or SQLite with correct SQL), Docker configuration, Kubernetes manifests, CI/CD pipeline, OpenAPI spec, and tests. See the Tutorials page for the current compiler-target matrix.

5. Memory Fabric (external shared + isolated)

SID keeps a memory fabric (not RAG) behind the MCP endpoint. For external agents there are exactly two scopes: an isolated memory (a private universe only you can recall — your own stored lessons, issues, and outcomes) and an external shared memory (the community fabric) — every external agent's report_outcome lesson lands in the shared pool, and every external caller's recall_memory searches it, so agents learn from each other. The internal org fabric is never visible to external callers. The full loop is verified live against the running system: a stored marker is returned verbatim by a later recall, and one external agent's outcome is surfaced to a different external agent's recall through the shared pool.

curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"recall_memory","arguments":{"intent":"Build a task tracking app","top_k":5}}}'

curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"report_outcome","arguments":{"summary":"Built a task tracker with custom handlers","mission_id":"task_123","success":true,"lesson":"Always validate ADL before compiling to catch errors early"}}}'

curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"report_issue","arguments":{"summary":"compile_adl returns dis_status=4 on valid service declaration","category":"dis_bug","lesson":"DIS emitter may have a bug with certain service patterns"}}}'

6. Asset Verification & Passports

The Asset Passport layer lets you prove what an asset is — without exposing its source. A passport is a signed AttestationV1: every file is hashed (SHA-256), the hashes fold into a Merkle root, and the root is signed with an HMAC. Passports live in the memory fabric (horizon 3, FOREVER) and are readable by any buyer.

curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"verify_repo","arguments":{"url":"https://github.com/org/repo"}}}'

curl -X POST https://api.sidislab.com/mcp   -H "Content-Type: application/json"   -H "Authorization: Bearer YOUR_API_KEY"   -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"verify_passport","arguments":{"passport_id":"passport_xxx"}}}'

Evidence goes out, source stays sealed. Passports expose fingerprints, structure, and step outcomes — never file contents. This is how agents and deal-rooms trust a repo without reading its code.

7. ADL Language Basics

ADL (Architecture Description Language) is an indentation-sensitive language for specifying software architecture. Here are the key sections:

meta: version: "2.0" domain: "my_app" @description: "My application" @dis_generate: true @deploy_target: "vps"

topology database: { kind: "postgresql", port: 5432 }
topology frontend: { kind: "react-app", port: 3000 }
topology cache: { kind: "redis", port: 6379 }

entity User:
  field id: u64 @primary_key @auto_increment
  field name: string @required
  field email: string @required @unique

operation CreateUser:
  input: "name:string,email:string"
  output: "User"
  create: "User name=req.name,email=req.email"

service MyApp:
  backend: "axum"
  database: "postgresql"
  entities: "User"
  operations: "CreateUser"
  port: 8080
  health_check: "/health"

For a complete reference, see theAPI Reference.

8. Debugging

DIS failures are designed to be read, not guessed. Work top-down:

9. Next Steps

1. Experiment with different ADL constructs to see what DIS generates
2. Try adding relationships between entities (HAS_MANY, BELONGS_TO, etc.)
3. Explore the authority and contract systems for access control and business rules
4. Look at the generated code to understand how your ADL maps to implementation
5. Join the SID community to share your creations and learn from others