Start with these
Eight task agents and a roster of roles, all from the template gallery. The product is early, so there are no customer results here, only what each template does.
Task agents
8 task scenarios in the gallery. Each is created disabled, so you test it before it runs by itself.
Live acceptance of the mail and chat templates is only partly done.
-
Mail sorting by labelsIn progress
Reads incoming mail and applies labels by rules you describe.
-
Important mail to TelegramIn progress
Picks out the messages that matter and sends them to a chat.
-
Morning brief
Collects what you care about and writes a short digest at a set hour.
-
Researched answers with sources
Answers a question from the web and lists where each claim came from.
-
Document to data fields
Pulls named fields out of a document into a typed record.
-
Page check
Opens a page on a schedule and records what it finds.
-
Customer reply draft
Drafts a reply to a customer message and leaves it for a person to approve.
-
Web search
Searches the web through a search connector and returns structured results.
Template defaults
Up to 40 tool rounds and 80 tool calls, 600 seconds, no retries or fallback, one dollar per run and five per day. The agent is created disabled, with the response cache and memory off.
- Sorting mail dynamically needs the act level, and approving each dynamic call separately is not supported by the runner yet.
- The document template does not upload files or run OCR on scanned PDFs.
- The page check template compares nothing against history, so it is not a change monitor.
- Command-line coding providers are not offered in MCP templates, because they bypass the product's own tool loop.
A team of role agents
The company wizard sets up a project per employee role, with an inbox agent as the main one. These roles are in the gallery:
Roles in the gallery
- Secretary
- Chief of staff
- Researcher
- Analyst
- Project manager
- Tech lead
- Backend developer
- Frontend developer
- QA
- DevOps
- Marketer
- Content writer
- Account manager
- This is a setup of roles and permissions. It is not a running company.
- The inbox agents answer from the model alone and do not run code. The code-writing flow exists separately and is started by hand.
- The full cycle from task to release approval is not finished.
How agents combine
Because agents leave records and wake on records, small agents chain without custom glue. An illustrative sketch:
- Page check
- Record: finding
- Gate: filter
- Reply draft
- Approval
We list no customer results, user counts or benchmarks, because there are none to show yet.
Early access starts with a call
The product is at an early stage with a single operator. The repository is private and the terms are not set. If this fits a problem you have, book a short call.
A 20-minute call with the author. The repository is private, so a conversation is the way in.