Skip to main content
Experimental version.This is the latest in-development version of AI-Implement. Features may change without notice and behavior is not guaranteed. Switch to the latest stable version here.
When planningEnabled=true, AI-Implement runs planning before implementation. The planning agent reads your source and writes structured comments. The orchestrator posts them to the originating issue in Linear or Jira. Planning can use Codex while implementation uses Claude, or the reverse. Select the planning agent, model, account, and timeout in Agent settings per stage.

​
How the planning phase works

1

Orchestrator dispatches the planning workflow

The orchestrator dispatches claude-plan.yml for an eligible issue with its issue context. The filename stays the same when you select Codex.
2

The planning agent reads the codebase

The agent reads and searches the repository without changing source files. It writes planning comments under ai-output/comments/.Codex planning can read and search repository text and write these comments. It cannot run shell commands, delegate work, or use network tools.
3

AI-Implement posts the plan

The supplied PLANNING.md template asks for:
  • Implementation Map — approach, files, and constraints for the implementer.
  • Acceptance Bar — testable claims for the reviewer.
  • Risks & Open Questions — uncertainties and prerequisites.
See Planning comment format for the filenames and headers.Planning output does not automatically start subagents. Existing repositories keep their own prompt templates until you edit them.
4

Issue marker is updated

After the planning workflow finishes successfully, the orchestrator updates the marker on the issue:
The AI-Planning label is removed and Plan-Complete is added.
5

Implementation starts automatically

The orchestrator dispatches the implementation workflow immediately after planning succeeds.
The autoApprovePlans field on each project mapping is reserved for a future manual-approval gate but does not currently affect dispatch behavior — implementation always starts automatically after planning, regardless of this setting.
6

Implementation workflow injects planning context

The implementation run uses the map as reference context. Review uses the acceptance bar to check the result. Planning context does not override the issue or repository instructions.

​
Disabling the planning phase

To skip planning for a specific project mapping, set planningEnabled=false on that mapping in the admin UI at /admin. Implementation will dispatch directly without a preceding planning run.

​
Choosing a planning model

For a configured project, use the planning stage’s agent settings. Its selected model and elapsed timeout take precedence over prompt front matter. For a legacy project, set model: in PLANNING.md:
For a legacy Bedrock project, supply a compatible Bedrock model ID or inference-profile ARN. Codex uses an elapsed-time limit. Claude’s native turn cap is separate; see Timeouts and limits.

​
Customizing the planning prompt

PLANNING.md supplies the planning prompt for either agent. Sync preserves existing copies. Add repository context without removing the output instructions.
Point the planner at the directories where decisions matter, such as migrations or shared API contracts.