titan_get_run whenever a tool returns queued or running, and to page through results too large for a single response. It reads state only—it never starts new work and never consumes credits.
Required scope: mcp:runs:read
Parameters
string
required
The
run_id returned by titan_search, titan_fetch, titan_crawl, or titan_run_template. Must be a UUID—this is the Titan execution_id.boolean
default:"true"
Include result records. Set
false for a lightweight status check while a long run is in flight.string
Requested shape for returned records.
integer
default:"100"
Records per page, up to 1000.
integer
default:"0"
Records to skip. Combine with
limit to paginate.Example
Response
Response fields
results is populated only for terminal runs. While a run is queued or running, use progress and steps to report advancement rather than waiting for partial records.Polling
Poll with backoff rather than in a tight loop. Every call is a real request, and hammering it will not make the run finish sooner.
Set
include_results: false while polling and switch it on once the status is terminal—that keeps each poll small.
Pagination
Results aggregate across all steps of a run, in step order. Page through them withlimit and offset:
records_returned against records_available in usage to know when you have read everything.
Warnings you may see
steps_unavailable exists so an agent can distinguish “this run produced nothing” from “results could not be loaded right now”. Treat it as retryable, never as an empty result.
Ownership
You can only read runs owned by the same Titan user as your API key. An unknown or out-of-scoperun_id returns a not-found error rather than someone else’s data.
Using the run elsewhere
run_id is the Titan execution_id, so the same identifier works across the platform:
GET /api/v1/executions/{run_id}— full execution detailGET /api/v1/executions/{run_id}/results/export— export the complete result set- Analytics and billing endpoints accept it as
execution_id
Next steps
- Runs and results — the full lifecycle
- titan_crawl — the tool that most often needs polling