Free Solo limits, billing, and ownership
What the current free plan funds, what is still planned, and who reviews agent output.
What Free Solo includes today
Solo is free to start with no card and no personal AI API key. Agent work is funded by ACP's platform wallet under fair-use caps. Project and agent concurrency limits depend on your account settings; check them in the app. This is not a timed trial that automatically converts into a paid Solo subscription. Paid checkout is not live yet; use the Pricing page for current plan details.
Ticket creation and agent execution have separate limits. Creating a ticket does not promise an immediately available agent, unlimited model spend, or a fixed completion time. Provider availability, fair-use controls, repository access, and the selected workflow can affect when work starts or advances.
You own the product decision
ACP records a builder's committed result, test outcomes, review findings, and delivery state so you can inspect the work. You still choose what to request, approve the plan where required, examine consequential changes, and decide whether the delivered behavior meets your needs. An approved agent review is evidence from the pipeline, not a warranty that every edge case was tested.
PR behavior also depends on the workflow. A Feature can produce one draft-to-ready pull request, a standalone ticket can produce its own PR, and a direct Build or Change can land on main. Check the run's output before assuming there is a PR awaiting your action.
Keep access proportional to the task
Connect only repositories and accounts needed for the project. Do not include secrets in a goal or in an MCP prompt. If a run is blocked on authorization, resolve it through the workspace's intended account or provider settings rather than asking an agent to copy credentials into code. For sensitive delivery, review the diff and deployed behavior with the same care you would apply to a human contribution.