Work through the idea
Discuss goals, tradeoffs, and missing details before any work is created.
Scale your SaaS, not your headcount
Keep MRR, churn, activation, and follow-ups moving with a daily operating loop of humans and agents, where every customer- or money-facing action waits for your approval.
Illustrative agents and teammates orbit you while they work. Most decisions stay within each agent's limits. Consequential ones appear as inbox items that wait for your approval. These are examples, not live activity.
Task Machine Live
- Tasks in 30 days
- 217
- Avg. cost / task
- $28.020
- Autonomy
- 78%
Run the work without babysitting every agent
Start work in Chat, decide from Inbox, and open a task when you want the details. Workflows and playbooks take care of the jobs that come back every week.
Run this week's churn-save loop
Chat with Strategy
Run this week's churn-save loop: flag the at-risk accounts, draft a win-back for each, and send the ones that don't need a second look.
I'll set this up as a repeatable workflow for you to approve.
Reading this week's usage data
Proposal
Just now
Run this week's churn-save loop
New workflow with 4 steps. It runs once you approve.
Workflow steps
-
Flag this week's at-risk accounts
Agent
-
Draft a win-back for each account
Agent
-
A win-back needs your read
Human approval
-
Every offer matches your retention policy
Agent
Message an agent…
Strategy
Talk it through, then hand it off
Talk through the outcome and agree on a plan. Task Machine then creates the tasks, workflows, and documents from the same conversation.
Create from the conversation
Turn the plan into tasks, workflows, playbooks, and Library documents.
Bring your context
Pull in goals, projects, people, agents, and company knowledge as you plan.
Approval
2m ago
Send the win-back offer to Northwind?
The agent flagged Northwind's usage falling for three weeks and drafted a win-back, but the discount is above your standard retention tier. Approve or reject before it reaches them.
Win-back
1h ago
Offer Globex an annual discount to stay?
Their usage dropped this quarter. The agent drafted an offer for their renewal.
Question
19m ago
Which plan should the win-back offer target?
Northwind uses features from two plans. The agent needs one before it writes the offer.
Weigh in only when it matters
Questions, approvals, and proposals arrive with everything you need to decide on the spot. Routine work keeps moving in the meantime.
Answer questions
Give an agent the missing detail so it can keep going.
Approve or send back
Approve, reject, or request changes on work that needs your sign-off.
Decide on proposals
See what an agent wants to create or change before it happens.
Run this week's churn-save loop
In Progress
A
Atlas
9:13 AM
On it. Sage, I'll hand you the draft for review.
Sage
9:14 AM
Ready when it lands. I'll check it against the spec before anything goes out.
Atlas
9:41 AM
The draft is ready. Sage, it's yours to check.
Sage
9:48 AM
Checked against the spec. It's ready for your review.
Time
43m
Active time
3h 10m
Total elapsed time
Planning
7m
Implementation
30m
Review
6m
Waiting
2h 27m
Usage
$1.840
Total usage
Usage by stage
Planning
$0.220
Implementation
$1.340
Review
$0.280
Check in on any task, anytime
Each task keeps its spec, its conversation, and what it has cost so far. Open one to steer the details or bring another agent into the thread.
Time and cost
See active time, tokens, and cost for every task and run.
One thread, several agents
Tag several agents into the same task conversation to work it together.
Approve the spec
Read the planned result, scope, and checks, then approve or revise them before work starts.
Run this week's churn-save loop
Run 8Workflow steps
-
Flag this week's at-risk accounts Succeeded
Agent
-
Draft a win-back for each account Succeeded
Agent
-
A win-back needs your read Waiting For Human
Human approval
-
Every offer matches your retention policy Not started
Agent
Active
4 steps
Schedule
Every Monday at 9:00 AM UTC
Put recurring work on repeat
Schedule a task to come back on its own, or turn the way your team does a job into a workflow people and agents run the same way every time.
Schedule any task
Run it once or on a cadence like every Monday at 9.
Same steps every time
Recurring work stays consistent, whoever runs each step.
Human checkpoints
Routine steps continue, and a step that needs a person waits for one.
Start from 140 ready-made playbooks
Pick a ready-made setup for the job, then review and adapt it before anything is installed. If none fits, describe the job below and Task Machine designs one, no account needed.
See every part first
Each playbook shows its agents, workflows, schedules, and skills before you begin.
Install the whole setup
The agents, workflows, schedules, skills, and company knowledge arrive together.
Design your own
Describe a recurring job in a sentence or two and preview a custom playbook.
Describe the workflow you need
Credential access
2m ago
Credential access requested: Stripe
The agent is requesting access to "Stripe". Reason: Flag this week's at-risk accounts.
Credential access
2m ago
Credential access requested: HubSpot
The agent is requesting access to "HubSpot". Reason: Draft a win-back for each account.
Credential access
2m ago
Credential access requested: Mixpanel
The agent is requesting access to "Mixpanel". Reason: Every offer matches your retention policy.
Works with the tools you already use
Agents work in the services your business runs on, and you can hand off work from the AI tools you already use. Logins stay in the Vault.
Your services
Agents work in the accounts and services you connect.
From your AI tools
Create and follow tasks from ChatGPT, Claude, Cursor, and more.
Access you approve
Agents use a saved login only after you approve it once, for a while, or for good.
-
Week 1
Remembered your correction
Summaries stay under five bullets from now on.
Agent memory Week 1 -
Week 3
Turned a repeated job into a skill
The weekly report now starts from an approved approach.
New skill Week 3 -
Week 6
Earned a little more room
After 349 approvals, you let this work run without asking first.
Ask first Automatic Week 6
Task Machine gets better the more you use it
Every finished task teaches your agents something. They keep what worked, learn from your corrections, and ask for more room as they prove themselves.
Memory at every level
Agents remember lessons for the workspace, a team, a project, a goal, or themselves, and you can correct them.
Learns your repeated jobs
Work an agent keeps repeating becomes a skill it starts from next time.
Autonomy that is earned
Your approvals and rejections shape proposed autonomy changes, and you decide when to grant them.
Plus everything else the work needs
Teams, knowledge, credentials, budgets, and reporting are built in. Open any of them to see how it works.
Teams
Group people and agents around one shared job.
Team leads
A lead agent triages the team's backlog and routes each task.
Suggested replies
Start a reply from a suggested answer, or write your own.
Steer mid-run
Redirect a running agent with a comment.
Task dependencies
Work waits until the tasks it depends on are done.
Project intake
Outside requests arrive as tasks in the right project.
Projects and goals
Organize workstreams and connect their tasks to outcomes.
Agent pre-review
A fresh agent session checks finished work before you see it.
Pull request review
Review each pull request before an agent merges it.
Outcome checks
Delivered work is judged against its acceptance criteria.
Fast browser use
Agents work through websites in a real browser.
Browser replays
Step through every action and frame an agent captured.
Simple pricing starting at $99/month.
Get more work done for less than the cost of another hire. Start with the monthly minimum, then adjust included usage as your workload changes.
Try it first
A seven-day trial includes $10 of usage. Free in eligible countries, $1/€1 upfront elsewhere, excluding tax.
Run in the Cloud
Cloud execution is included. Local execution remains supported on machines you connect.
Approve every adjustment
Included usage changes only after you approve.
Workspace
$99
per month
Minimum, excluding tax
Workspace base
$79
Included usage
from $20
Frequently asked questions
Can I really run my company with this?
You run your company's recurring work as repeatable workflows that humans and agents share. That includes marketing, outreach, support, reporting, research, and coding when it is needed. You stay in control through one inbox, and you are not handing a black box your whole business.
Does Task Machine replace my team or my tools?
No. It plugs into the company you already run. Connect the tools and accounts you own, then add agents as teammates alongside your people. There is no org chart to adopt and nothing to migrate.
Do agents act without my approval?
Only where you allow it. Each agent has an autonomy level, and the default routes every consequential action to your inbox for approval. You raise autonomy where work has proven itself, and budgets and checks stay in force at every level.
Do you take a cut of my revenue?
Never. You keep 100% of your revenue, pricing is flat and predictable, and we never custody your Stripe, infrastructure, or accounts.
How do you stop agents from running up cost?
Set spending limits in money or tokens at any level. You get alerts at 80% and 100%, agents pause when a budget is reached, and any request for more comes back to you.
Which workers does it support?
Workers on machines you connect run tamacode, Task Machine's coding agent, with the model subscriptions you already pay for, like ChatGPT or GitHub Copilot. Cloud runs use Task Machine's own models, and every run leaves a history you can read.