world.read
Entity discovery
Resolve the entity or package before requesting a record.
Agent integration
The tools let an approved agent find an entity, read a record, fetch its sources, and see when access or evidence is missing. The contract is verified locally. Hosted access is not generally available.
Tool categories
world.read
Resolve the entity or package before requesting a record.
world.read
Find stale assumptions, missing evidence, and contradictions inside approved context.
world.read
Read records, recent changes, domain context, and visible blockers.
evidence.read
Fetch the source trail before citing a claim.
decision.read
Read typed records prepared for an approved workflow.
stream.read
Read updates for records and packages included in the caller's access.
simulate.run
Run a clearly labeled scenario when the caller has explicit access.
webhook.manage
Inspect delivery records. External webhook creation and customer secret rotation are not currently approved.
Agents query context, evidence, packages, and updates before considering any downstream workflow.
Sensitive capabilities require explicit authorization and may return redacted outputs.
Responses preserve enough context for a human, system, or auditor to understand what was used.
Scenarios are separate from observed records and require explicit access.
Connection shape
The configuration below shows the expected shape. It is not a production configuration. Approved clients receive a base URL, token, scopes, and package instructions for their specific use.
Illustrative MCP config
{
"mcpServers": {
"crowdalpha-agentlayer": {
"command": "python",
"args": ["-m", "services.agentlayer.mcp_server"],
"env": {
"CROWDALPHA_API_BASE": "BASE_URL_FROM_APPROVED_HANDOFF",
"CROWDALPHA_API_TOKEN": "TOKEN_FROM_APPROVED_HANDOFF"
}
}
}
}Local verification
# local contract checks, not a hosted service probe python -m py_compile services/agentlayer/mcp_server.py python -m pytest services/agentlayer/test_mcp_contract.py -q pnpm agent:discovery-check