Flanner and Plannotator
Last updated: 23 September 2026
Plannotator helps teams review a plan before an agent executes it. Flanner manages the plan's entire operational life: capture, immutable history, code anchoring, freshness, authorization, team availability, and agent retrieval.
The names sound like rivals. In practice the two tools act at different moments in a plan's life, and plenty of people will want both. Here is how they differ, and how they fit together.
At a glance
| Plannotator | Flanner | |
|---|---|---|
| What it is for | Reviewing an agent's plan or code in your browser before it runs, and sending anchored feedback back to the agent. | Keeping the plans an agent saves: every version, checked against the code, shared with your team. |
| When it acts | At approval. In Claude Code, a plugin hooks the moment Claude asks you to approve its plan. | Afterwards. Your agent saves a plan through flanner's MCP tools, and each revision becomes a version. |
| Agents | Claude Code, Codex (through an experimental hook, not on Windows), OpenCode, Pi, Copilot CLI, Gemini CLI, Kiro CLI and Amp. Droid by hand. | Claude Code, Codex and Claude Desktop, set up by flanner init. Any MCP client can call its tools. |
| Where plans live | ~/.plannotator, in your home folder. | .plans/ in the repository, kept out of git until you decide otherwise. |
| Version history | Yes. Each revision is saved and can be compared, grouped by project, heading and day. | Yes. A numbered file per version, read with flanner history and compared with flanner diff. |
| Checks a plan against later code | Not documented. | flanner freshness gives each plan a status, from fresh to stale, with the evidence behind it. |
| Code review | Local changes, GitHub pull requests and GitLab merge requests. | Not a code review tool. |
| Sharing with a teammate | A link. A small plan travels inside the link; a larger one is encrypted in the browser and uploaded for seven days. Hosted Workspaces are in private beta. | Flanner Mesh, device to device and signed. Its servers manage accounts and access, and never receive plans. |
| Licence and price | Apache 2.0 or MIT, free. No price published for Workspaces. | MIT, free. Mesh from $9 a month. |
Where they differ
The moment they act
Plannotator sits at the approval step. When Claude Code finishes a plan, Plannotator opens it in your browser, you mark it up, and your notes go back to the agent as structured feedback. It does the same for code changes and pull requests. Flanner comes in once a plan is worth keeping: it gives the plan a home in the repository and records each revision after that.
Whether the plan still matches the code
Both keep a plan's earlier versions. Flanner also reads each plan against your git history: a plan that cites a file which no longer exists, or code that has seen dozens of commits since, is flagged with the evidence. Plannotator's docs describe comparing one version of a plan with another, not a plan with the code.
How a teammate sees it
Plannotator shares a review as a link, and its hosted Workspaces are where it is taking team features. Flanner Mesh moves plans between your team's own machines, signed by the device that wrote them and verified on arrival. It is a paid tier, and the plans never pass through flanner's servers.
Using both
- Let Plannotator catch the plan at approval, and review it there with your notes.
- Once you approve, ask your agent to save the plan through flanner. The guidance flanner writes into
CLAUDE.mdandAGENTS.mdtells it how. - From then on, each revision is a flanner version, and
flanner freshnesstells you when the code has moved away from it.
The two do not step on each other. Plannotator hooks the approval prompt; flanner's only hook guards direct writes into .plans/.
Which to choose
- You want to review plans and code diffs in a proper interface before an agent acts: Plannotator.
- You want every plan kept, told when one has gone stale, and shared across your team's machines: flanner.
Version control for Claude Code and Codex plans covers keeping plans with git alone, and the full comparison sets flanner beside the agents' own features.
Sources
Checked on 23 September 2026. If something here has changed, tell us and we will correct it.