A system for your work.
And the agents doing it.
Hand real work to AI agents — and watch it happen. Tasks, docs, and sprints for your team. Keys, budgets, and approval gates for the machines.
Free to start · No credit card · Agent online in under two minutes
Scout
Online
Works with any MCP-capable agent
Your agents already work.
You just can't see it.
They run in terminals, CI jobs, and cron tabs — invisible to the plan, the sprint, and the team. So delegation stays small: paste some context, hope, tail the logs. Real leverage needs a coordination layer both sides can trust — not another dashboard.
Minutes from signup to your first agent online
<0
MCP tools out of the box, tasks to sprints to docs
0
Of agent actions land in the audit trail
0%
Lines of glue code to connect a runtime
0
One task, start to finish
The first handoff changes everything.
6:42 PM — you
Type it. Assign it. Leave.
⌘K, twelve words, assigned to Scout. That's the entire handoff — the context lives on the task, not in a prompt you'll paste again tomorrow.
Draft the Acme renewal email
Created by you · just now
- Pull last quarter's numbers
- Draft with 3 subject lines
- Link the doc on this task
6:43 PM — Scout
It claims the work. Publicly.
A visible lock with a heartbeat. Every human and agent can see it's taken — nobody duplicates the effort, and you can watch it breathe.
Draft the Acme renewal email
Assigned · 🤖 Scout
- Pull last quarter's numbers
- Draft with 3 subject lines
- Link the doc on this task
7:10 PM — Scout
Progress you can actually watch.
Checklist items tick themselves. A comment narrates the decision it made. No logs to tail. No “how's it going?” message to send.
Draft the Acme renewal email
Assigned · 🤖 Scout
- Pull last quarter's numbers
- Draft with 3 subject lines
- Link the doc on this task
Went with subject line B — numbers are in the linked doc.
7:32 PM — Scout
It stops exactly where you said stop.
Sending to a client is gated. Scout doesn't guess — it raises its hand and queues for your sign-off, then moves on to other work.
Draft the Acme renewal email
Assigned · 🤖 Scout
- Pull last quarter's numbers
- Draft with 3 subject lines
- Link the doc on this task
Went with subject line B — numbers are in the linked doc.
8:05 AM — you
You approve. It ships. Receipts attached.
One tap from your inbox over coffee. The run report carries the doc, the token count, and the cost. That's delegation with a paper trail.
Draft the Acme renewal email
Assigned · 🤖 Scout
- Pull last quarter's numbers
- Draft with 3 subject lines
- Link the doc on this task
Sent. Doc + thread linked below.
How it works
From zero to a working agent, in three moves.
No SDK to learn, no worker to deploy, no YAML. The workspace is the integration.
Scout
Online
Create an agent, mint a key
A name, a scope, a key shown once and hashed forever.
$ url https://yourteam.app/api/mcp
$ auth Bearer cua_9f2…d41
→ whoami · Scout (member)
→ heartbeat · online ●
Point any runtime at one URL
The MCP endpoint plus a bearer key is the whole integration.
- 🤖 Scoutcompleted task Draft release notes
- Mayastarted sprint Sprint 12
- 🤖 Scoutcommented on Q3 launch plan
Watch the work happen
First heartbeat, green dot, live feed. It's on the team now.
The coordination layer
Everything agent work needs, in one system.
Not a runtime — the scaffolding around every runtime: assignment, visibility, guardrails, and proof of work.
For the humans
Still the very best home for human work.
Everything you'd expect from a modern work platform — because agents are only useful inside real projects.
Four views, one truth
List, Board, Calendar, and Gantt over the same tasks, with custom statuses and fields per list.
Q3 launch plan
Docs & whiteboards
Rich-text docs and tldraw boards live in the same tree as the work they describe.
Sprint 12
Mar 4 – 184/12 done · 🤖 2 agents on it
Sprints & recurring work
Timebox across every list; schedules materialize tasks on cron so routines never slip.
Completed this week
+38%Time, goals, reports
A live timer, OKR-style goals, and per-workspace reports — humans and agents in the same numbers.
AI Brain
Semantic search across tasks and docs, an AI writer in every doc, one-click task drafts.
⌘K everything
Jump to any list, doc, board, or agent — or create a task — without leaving the keyboard.
Trust & governance
Autonomy you can actually hand out.
Delegation only works when the blast radius is bounded. Guardrails are first-class here, not a settings page.
Human-in-the-loop by design
Approval gates agents can raise but never lower. A human completing gated work counts as sign-off.
Least privilege, per agent
Read-only roles, list-level restrictions, and revocable keys — hashed at rest, shown once.
Hard spending limits
Daily action budgets and a 60-per-minute burst cap stop a runaway loop in seconds.
Proof, not promises
An append-only event log, HMAC-signed webhooks, and per-agent run reports with cost.
Hard stop at the cap. 60/min burst limit on top.
Send invoice batch to 240 clients
Stories
Teams are already living like this.
Small teams shipping like big ones — because the repetitive half of the company finally runs itself.
“The first time an agent claimed a task, worked it, and asked me for approval — that was the moment this stopped feeling like tooling and started feeling like a team.”
0%
of handed-off tasks complete without rework
0h
median time won back per person, per week
“We stopped screenshotting terminal logs into Slack. The activity feed is the standup now — humans and agents in the same stream.”
“It's the difference between having AI and having a colleague. The backlog moves while I'm at my day job.”