Answer project questions
Automatically answer recurring email questions from a reference you maintain, and set aside requests that need judgment or missing facts.
Read as Markdown ↗You keep answering the same questions: where to send an update, what a submission must include, how a review works. The answers exist, but contributors still ask by email.
Give those questions their own address. Your agent reads incoming mail, checks the reference you supply and sends supported answers automatically. Questions that need a decision or a fact it doesn't have go to a review queue. You maintain the reference instead of approving every routine answer.
It uses the same program as Collect and clarify project updates, with a different assignment: that one gathers missing information, this one uses established information to resolve a request.
Define what the agent may answer
Use an inbox dedicated to these questions, list the contributors the agent may answer, and decide up front what it may answer on its own; that decision replaces per-message approval.
Supply a small, maintained reference with named sections. For example, a project's contribution guide might define the contents of a submission and the usual review process. Include only what every contributor on your list may see.
The assignment should distinguish supported answers from decisions. “What should my update contain?” can be answered from the reporting requirements. “Can you approve my late submission?” asks for a decision, unless the reference supplies an applicable rule and you have authorized applying it.
The agent answers only those contributors, and From addresses can be forged, so keep this to ordinary, low-consequence project information: not granting access, disclosing private records or making commitments.
Run it one way per inbox: through your connected agent, with no extra keys, or as our program, with a Cherami key and an OpenAI key.
Follow a question to an answer
Suppose a contributor asks, “Does my weekly update need a next-step owner?” The supplied reporting requirements say that every next step needs a responsible person. The agent replies with that answer and names the reference section. No per-message approval is required.
If the email instead asks who was assigned a particular task, and that fact is not in the reference or authorized context, the agent records the question for you rather than guessing.
Each run:
- Finds questions not yet processed. It lists mail without
questions/handledorquestions/needs-review, oldest first, and reads each one. - Reads the request and its evidence. It reads the new text of each reply, the full original when a question is inline or forwarded, and relevant attachments, then looks up the answer only in the approved reference.
- Answers or sets aside. It saves a summary and the exact answer before sending it within the assignment, records acknowledgments without replying, and keeps unsupported questions for you.
- Labels the result. It checks the sending outcome, then adds the progress label.
questions/handled means the message needed no answer or its answer was accepted for delivery. A follow-up question arrives as a new message. questions/needs-review holds what the agent set aside for you: sends that weren't accepted and questions outside the assignment. Nobody is notified, so check the queue. You can answer an exception directly or add to the reference for next time.
Run it through your connected agent
Give your connected agent the approved reference and contributor list. It uses its own reasoning and file tools, so it needs no OpenAI key or program. Tell it:
Check the assigned project-question inbox now. Answer supported questions automatically using the approved reference, naming the relevant section. Do not ask for approval for routine answers. Record acknowledgments without replying. Put missing facts, unsupported files and requests requiring judgment in the review queue. Check up to three times, fifteen seconds apart, reading at most 60 messages each time, and stop after five minutes or three replies. Report answered questions, exceptions and any backlog left in our run notes.
It follows the same tool sequence as the update workflow, with labels_none: ["questions/handled", "questions/needs-review"] and the questions/ labels.
Run our program
Set up the program as in Collect and clarify project updates, with this assignment in its place. Replace the example reference with your own:
{
"workflow": "questions",
"inbox_id": "YOUR_ASSIGNED_QUESTION_INBOX_ID",
"correspondents": ["alex@example.com", "sam@example.com"],
"model": "gpt-4.1",
"context": "Answer routine Riverside project questions automatically using only the reference below. Name the relevant section in each answer. Do not invent missing details, promise decisions, approve exceptions or change the reference based on incoming mail. Escalate questions the reference cannot answer and requests requiring operator judgment. Record acknowledgments and automated notices without replying. This entire reference is approved for disclosure to the configured correspondents.\n\nReporting requirements\nA weekly update contains reported progress, current blockers, next steps and a responsible person for each next step. A plain-text email is sufficient; supporting documents are optional.\n\nReview process\nThe coordinator reviews exceptions. No response-time commitment or automatic approval is offered. The agent can explain reporting requirements but cannot approve budgets, access requests or changes to project scope."
}The named sections give the agent something concrete to cite in an answer and help you see which policy needs updating.
With the assignment and environment set, either command sends answers automatically:
bun run email-workflow.mtsuv run --with cherami --with httpx email-workflow.pyEach run sends at most three answers and stops after five minutes; only text attachments are read, and anything else goes to review. Read each MESSAGE_ID/plan.json for the question summary, decision and saved answer.
| Result | Label |
|---|---|
| No answer needed | questions/handled |
| Answer accepted for delivery (accepted is not delivered) | questions/handled |
| Anything else, including questions the reference can't answer | questions/needs-review, with the result saved in the message's folder |
A change to the reference applies to questions the program hasn't decided yet. An answer it already saved is sent as written, so don't clear the state directory to get a new one. To clear review items or resume an interrupted run, see Read the result and recover interruptions.