​
What agents can do

Agents use Orgs as a legal and governance boundary for organization-level work. The agent can draft, request, propose, and execute only within the credentials and policy it has been granted. Common tasks:
  • Create an entity from a recipe or constitution.
  • Create and inspect governance proposals.
  • Read treasury and compliance state.
  • Request actions that require human approval.
  • Retrieve audit records for lineage and accountability.

​
Production API

Use the production API base URL: 
export ORGS_API_URL=https://api.orgs.sh
export ORGS_API_KEY=sk_live_...

curl -sS "$ORGS_API_URL/health"
Store credentials in your runtime secret manager. Do not put API keys in prompts, source code, logs, or long-term agent memory.

​
MCP configuration

For MCP-compatible runtimes, expose Orgs as a tool server: 
{
  "mcpServers": {
    "orgs": {
      "command": "orgs-mcp",
      "env": {
        "ORGS_API_URL": "https://api.orgs.sh",
        "ORGS_API_KEY": "sk_live_..."
      }
    }
  }
}

​
Operating rules

Agents must follow the constitution, not only the API schema.
  • Read entity state immediately before mutating it.
  • Create a proposal when risk is unclear.
  • Never bypass spend, contract, hiring, or amendment thresholds.
  • Surface legal, filing, treasury, or compliance failures to a human operator.
  • Preserve the user intent that caused each entity-level action.

​
Suggested permission boundary

Start with read-only permissions, then grant write scopes as the agent demonstrates need:
  1. Read entity, proposal, treasury, and audit state.
  2. Draft proposals without execution authority.
  3. Submit formation requests from approved recipes.
  4. Execute actions only after proposal approval.

​
Next steps