auth.md, and it can authenticate, find endpoints, and
call them. The setup is the same in any agent framework or coding assistant.
Prerequisite: create an API key
Create a key in the dashboard and give it a name that identifies the agent, such assupport-triage-agent. Set a
credit limit on the key so one agent
can’t spend the whole balance.
Store the key as ENCRATA_API_KEY where the agent runs:
Point your agent at auth.md
https://encrata.com/auth.md tells an agent how to authenticate, how to verify
its key with the free GET /api/account/credits call, and where to find each
endpoint. Add it to the agent’s instructions:
Instructions for your agent
Coding agents
Build your integration with a coding assistant that readsauth.md, writes the
code, and runs it.
Cursor
Point Cursor at auth.md.
Claude Code
Build and test from the terminal.
Windsurf
Point Cascade at auth.md.
Docs for agents
Every page in these docs is available as plain text for agents:Add Encrata as an agent tool
A tool is a function your agent can choose to call. Write it once in plain Python, then register it with your framework’s tool wrapper, such as LangChain’s@tool, CrewAI’s @tool, or LlamaIndex’s
FunctionTool.from_defaults.
- The docstring is the tool description. The model reads it to decide when to call the tool and how to read the result, so list the real values the endpoint returns and what the call costs.
- Return errors instead of raising them. A
402or429that raises an exception ends the agent run. A returnedmessagelets the agent tell the user what happened. See Errors and Rate limits. - Return only the fields the agent needs. Smaller results cost fewer tokens and are easier for the model to reason about.
Test with safe inputs
Use inputs that don’t belong to real people while you build and test.Next steps
auth.md for developers
Key management and system prompts for agents.
Use cases
Real jobs to give your agent, with the fields to check.
Lookups
Every lookup, its input, and its cost.
Credits
What each lookup costs.