Switch from aiagent.app to Task Machine

Moving recurring work between platforms is not a file transfer. The difficult part is preserving context, credentials, control points, ownership, and evidence without interrupting the live job.

aiagent.app and Task Machine share agents, skills, memory, knowledge, visual Workflows, schedules, approvals, Inbox review, project work, and run history. Migration is still a manual rebuild. Task Machine does not import aiagent.app agents, Company Brain data, Workflow graphs, run history, Audit Log events, or Board tickets.

Decide which recurring jobs should survive, then recreate their context, execution, controls, and ownership in Task Machine. Keep the source Workspace available while validating each replacement, and export or copy any records required by your retention policy before cancelling the old plan.

Why do people switch from aiagent.app?

• Start with the job. aiagent.app makes agents, squads, knowledge, connections, and workflow blocks available for composition. Task Machine Playbooks start from a named recurring outcome and install the Agents, Workflow, Skills, Documents, and setup values for that job.
• Keep humans and Agents in one accountability model. aiagent.app has teammates, assignment, Board tickets, and approvals. Task Machine makes human and Agent members share the same Projects, Tasks, comments, ownership, and handoffs from the start.
• Separate generation from verification. aiagent.app exposes scorecards, performance views, traces, and approval nodes. Task Machine also puts verifier nodes directly in the production Workflow, so a failed acceptance check stops the run and becomes an actionable Inbox decision.
• Run supported coding work in your environment. aiagent.app's public product is hosted and connects to approved tools. An optional Task Machine Local Worker can run supported coding-agent tools beside local repositories, files, browsers, and CLIs.
• Pool operational usage. aiagent.app publishes tiered plans and a per-member Teams offer. Task Machine combines Workspace access with recurring included usage pooled across the Workspace, then meters additional usage.

You keep 100% of your revenue, Task Machine takes no cut, and Task Machine never custodies your accounts.

What maps to what?

The mapping is conceptual. Node types, data shapes, credentials, and histories do not transfer automatically.

What should you inventory first?

Begin with active recurring work rather than every object in the old Workspace. For each Autopilot or workflow, record:

1. the business outcome and current owner
2. its trigger or schedule
3. the agent roles and instructions involved
4. the knowledge, memory, and skills each role needs
5. every connected account and concrete action
6. branches, waits, approvals, and retry behavior
7. the evidence used to accept or reject the result
8. the Board ticket, Inbox item, or report produced at the end

Use this inventory as the replacement's acceptance checklist. Leave old experiments, unused agents, and duplicated context behind.

How does the switch work?

1. Create the Workspace and members. Invite the people who will own work, then create the Agent members required by the first recurring job.
2. Install the closest Playbook. Use its preview and setup questions to provision a starting Project, Workflow, Agents, Skills, and Documents. If no Playbook matches, build one Workflow around the inventoried process rather than reproducing the whole aiagent.app Workspace.
3. Move trusted context deliberately. Copy current policies and source documents into the Library. Attach only the material each Agent needs. Do not treat a transcript dump or old run history as clean knowledge.
4. Reconnect execution. Create the relevant Connectors and credentials. Use Managed Cloud by default, or install a Local Worker when supported coding work needs local repositories, files, browsers, or CLIs.
5. Rebuild the control path. Map starts, agent steps, tools, branches, waits, and approvals. Add verifier nodes wherever the acceptance checklist can be evaluated independently from the Agent that created the output.
6. Run one representative case. Keep autonomy conservative. Compare the result, evidence, approval behavior, side effects, and completion state with the source workflow.
7. Schedule only after validation. Enable recurring execution after the representative run passes and the responsible human knows which decisions will arrive in Inbox.
8. Retire the source job last. Leave the aiagent.app Autopilot disabled but available until the replacement has completed on its real schedule and any required records have been retained.

What do you give up?

aiagent.app presents more of the agent system in one hosted visual product. Its product tour places Company Brain, the model catalog, tables, visual builder, squads, and aggregate Performance in one navigation model. Paid users can bring provider keys, and teams can buy a managed service instead of doing the implementation themselves.

Task Machine asks you to choose or shape a Playbook, connect accounts, and operate a Worker when local coding work requires one. Its explicit Task and verifier model creates more structure. If your current aiagent.app setup already runs reliably, its hosted execution boundary suits your requirements, and its performance views answer the questions you care about, migration may add work without enough benefit.

What should you know before switching?

Can Task Machine import my aiagent.app company automatically?

No. There is no automated import for agents, squads, Company Brain, workflows, tickets, audit history, or performance data. Use the mapping and inventory above to rebuild one recurring job at a time.

Should I recreate every aiagent.app agent?

No. Create the Agent roles needed by the first validated Workflow. A Playbook may combine responsibilities differently, and unused experiments do not need a destination.

What replaces an aiagent.app scorecard?

Use acceptance criteria and a verifier node when a production run must pass an independent check before continuing. Run history and usage records support investigation. This does not reproduce every aggregate in aiagent.app's Performance view.

Can I keep using aiagent.app for some work?

Yes. A staged migration limits cutover risk. You can keep aiagent.app for hosted agents or its company-level performance view while moving a job that needs shared Task accountability, explicit verifier gates, or supported coding tools on a Local Worker.

For the evaluation view before planning a move, read Task Machine vs aiagent.app.