AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
Conductor Send Message
Sends a prompt to a Conductor agent session and, by default, waits for the agent to finish and returns its reply. Requires your own Conductor API key, created at app.conductor.build/users/api-keys and added through AutoGPT’s credentials UI. Select that credential for each Conductor block; no server-wide default key is used.
Conductor Send Message
What it is
Send a prompt to a Conductor agent session and, by default, wait for the agent to finish and return its reply.
How it works
The block posts {message} to POST /v0/sessions/{id}/messages and returns the receipt (message_id, state queued or sent, deep_link). With wait_for_reply (the default) it then polls every poll_interval_seconds: each poll reads the transcript rows that arrived since the previous one and then GET /v0/sessions/{id}/status, so the status is never older than the rows it is judged against. The receipt ID is the prompt's content.id in the transcript, not its row ID, and the API only accepts row IDs as after cursors; the block looks for that row among the newest 300 rows, then in up to 1000 older rows, and otherwise resolves the turn from agent rows tagged with the receipt as their turnId, reporting the omitted history through truncated. Either way it returns the prompt row's ID as next_after, ready to pass to Get Session's after (without wait_for_reply it checks the newest rows once; while the prompt is still queued and has no row, next_after is the receipt itself, which Get Session also resolves). It finishes when the session reports error, or idle after the turn has progressed past its startup events (Claude system/command_lifecycle, Codex thread.started/turn.started), reading the transcript once more for rows written just before the status changed; a session that is idle because the prompt is still queued, or has only launched the agent, is not treated as finished. reply joins the visible agent text of the turn (Claude assistant text, completed Codex agentMessage items) and message_count is the number of rows of the turn that were kept (the newest 1000 when a turn is longer; truncated is set). The raw rows are only returned, in messages, when include_messages is on. timeout_seconds is a wall-clock bound: sleeps and requests are capped by it, nothing is requested once it has elapsed, and when it elapses timed_out is set and whatever arrived so far is returned. The block's own execution cap is two hours, so timeout_seconds is limited to that.
Inputs
| Input | Description | Type | Required |
|---|---|---|---|
| session_id | Session to send the prompt to | str | Yes |
| message | Prompt for the agent | str | Yes |
| wait_for_reply | Wait until the agent is idle and return its reply | bool | No |
| timeout_seconds | How long to wait for the reply | int | No |
| poll_interval_seconds | Seconds between status checks while waiting | int | No |
| include_messages | Also return the raw transcript rows of the turn in messages. Off by default: a long turn is hundreds of kilobytes, while reply, session_status and next_after cover the usual needs | bool | No |
Outputs
| Output | Description | Type |
|---|---|---|
| error | Error message if the operation failed | str |
| message_id | ID of the sent prompt | str |
| state | queued or sent | str |
| deep_link | Link that opens the message | str |
| next_after | Transcript row ID of the prompt's row; pass it as after to Get Session to read the agent's turn. Falls back to message_id while the prompt has no row yet, which Get Session also accepts |
str |
| session_status | idle, working or error once waiting finished | str |
| reply | Text the agent produced in response | str |
| message_count | Number of transcript rows of the turn that were kept | int |
| messages | Raw transcript rows of the turn, oldest first; only emitted when include_messages is on | List[Dict[str, Any]] |
| timed_out | True when the wait ended before the agent went idle | bool |
| truncated | True when the turn produced more rows than are kept; the newest ones were kept and reply may be incomplete | bool |
| error_message | Session error, if any | str |
Possible use case
Conversational driver: Let AutoPilot steer a coding agent turn by turn: send a prompt, read reply, decide the next instruction.
Fire and forget: Turn wait_for_reply off to queue several prompts and check back later with Get Session.