Bring your own agent
Your agent. Your model.
Our board.
Track40 does not bolt on an AI and charge you credits for it. Give the agent your team already uses a token, and it works your pipes alongside everyone else: reading jobs, filling in fields, moving cards and leaving notes, under exactly the same rules.
No AI credits. No AI add-on. No markup on your model.
What your agent can do
Read the board
List, search and read cards with names already resolved, so the agent sees “Phase: Awaiting approval” and “Assignee: Kim”, not a column of ids.
Do the paperwork
Create cards, fill in fields, set labels and due dates, and move a job to its next phase once it has what it needs.
Leave a note
Post a comment on a card for the next person, such as a summary of the customer’s email or the quote it drafted.
Hear about new work
Webhooks tell your agent the moment a card is created, moved or commented on, with the whole card in the payload.
Answer a button
Put an action button on a card. When someone presses it, your agent gets a signed request with the card attached.
Everything else
The board stays the place your people work. The agent takes the repetitive steps, and the humans keep the decisions.
A safe place to let an agent work
The hard part of putting an agent on real work is trusting it. These are the limits Track40 holds it to, whatever model is behind it.
Tokens with only the access it needs
Mint a token that can only read cards, or only comment, or only write cards. A restricted token is refused everywhere else. Revoke it in one click.
It cannot skip a required field
Agent writes go through the same validation as a person in the app: required fields, formats, cross-field rules and the phase moves you allow. A rejected write comes back listing which fields failed, so the agent can fix it and try again.
It sees what its owner sees
A token acts as the person who made it. Pipes, cards and fields that person cannot see are stripped out before anything leaves the server.
Every change is on the record
The activity feed shows what changed and when, under the token’s owner. A deleted card can be restored, so a bad run is not a lost afternoon.
Spending has a ceiling
Agent calls draw the team’s API allowance first, then a prepaid wallet. Auto top-up stops after 2 charges a month, so a runaway loop cannot keep billing your card. Reads keep working even when the wallet is empty.
Webhooks you can trust
Every delivery is signed with HMAC-SHA256 and carries an event id that stays the same across retries, so your agent can check where a request came from and never act on one twice.
Connect one in three steps
- 1
Mint a scoped token
In the app, under Settings, then Tokens. Tick only the scopes the agent needs. The token is shown once, so store it like a password.
- 2
Give your agent the API reference
The API reference is one plain page of endpoints and examples. Hand it to your agent with the token and it has everything it needs.
- 3
Add a webhook so it hears about work
Point a pipe’s webhook at your agent and it wakes up when a card lands, instead of checking the board on a timer.
T40=https://app.track40.com/api/teams/ac3me
# What is waiting for a quote?
curl "$T40/pipes/sq4mx2ph/cards?phaseId=ph2qzk4n" \
-H "Authorization: Bearer $TOKEN"
# Fill in the quote and move the job on
curl -X PATCH "$T40/cards/3f1c9b2e-…" \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{"values":{"qt7m2xfa":1850},
"phaseId":"ph8wm3xs"}'| Scope | Lets the agent |
|---|---|
cards:read | Read pipes, cards, comments, members and attachments |
cards:write | Create, update, move, delete and restore cards |
comments:write | Post, edit and delete comments |
automations:write | Manage a pipe’s webhooks |
What it costs: no AI credits
Your model, your billYou pay your AI provider directly at their price. Track40 adds nothing on top.
5,000 calls / editor / monthAgent requests count as ordinary API calls, pooled across the team. Five editors share 25,000 a month.
Then US$1 per 1,000Busy months draw a prepaid wallet you top up yourself, with a cap on automatic top-ups.
Questions
- Which agents work with it?
- Anything that can make an HTTPS request with a bearer token. That covers coding agents with a terminal, agent frameworks with an HTTP tool, scripts your team already runs, and workflow tools such as Zapier and Make.
- Why no built-in AI?
- Because your team picks its own model and its own agent, and those choices change every few months. We would rather make the board easy for any of them to work than sell you credits for one we chose.
- Does Track40 send our data to an AI provider?
- No. Track40 has no AI of its own. Your agent receives only what its token can read, and what happens next is between you and the provider you chose.
- Can the agent’s work show under its own name?
- A token acts as the member who created it. If you want the agent’s changes to read as its own, invite it as a member with its own login and mint the token from that account. Making changes takes an editor seat, like any other member.
Put your agent on a real pipe
30 days free, no card required. Set up one workflow, mint a token, and see what your agent can take off your team’s plate.