Human guide: work with your agent →
Connect Deepline as a remote MCP server and verify authorization before running
a lookup. Your first task is tool discovery in the intended workspace.
Before you start
Use an MCP client that supports remote servers and Deepline’s authorization
flow. Shell-capable agents can instead use CLI setup.
1. Add the remote server
Use your client’s custom connector or remote-server setup. Supply:
Complete Deepline sign-in and select the workspace the client should use.
Do not copy a provider API key into the OAuth fields.
Client screens vary. For Claude, consult its
remote connector instructions.
The remote server must be reachable from the client’s execution infrastructure.
2. Test discovery before execution
Ask the client:
Read the returned identifier and required inputs. A plausible tool name is not
a valid payload. Search results do not establish authorization for every action.
Deepline discovery works, the intended workspace is authorized, and the
selected contract accepts the identifiers you plan to supply.
3. Run one bounded task after review
Use one known person or a small reviewed sample. Ask for source evidence,
validation outcomes, unresolved results, and actual Deepline usage.
Use Work email to choose the right input path.
MCP access does not change the Play’s schema or create a universal fallback
route. The same workspace and action contracts still apply.
If connection or authorization fails
Check the remote URL and completed authorization. Confirm the workspace rather
than repeatedly adding connectors. Distinguish network reachability from
provider access and no-result outcomes.
Ask your client to explain the specific failed boundary before retrying.
For recurring custom work, continue with Write Plays. 
