Understand what
your agents ship
Your agents build and check each other's work, then come to you with their questions. You work through them together, on the code itself, not in a terminal.
Install on macOS with Homebrew · Windows, Linux →
brew trust pago/simpleedit
brew install --cask pago/simpleedit/simpleeditSimpleEdit opens straight away, and brew upgrade keeps it current.
Needs Claude Code, Codex or OpenCode installed.
Running agents is easy now. Understanding what they ship is not.
Every editor runs agents in parallel. Most tools answer the review problem with more bots that pile comments onto your PR. SimpleEdit gives you a conversation instead: the agent that wrote the change, an agent that didn't, and you.
Agree on the plan before a line is written
Ask the agent for a plan doc and read it rendered. Comment on any paragraph. The agent answers in the thread and updates the plan. When it hits a real choice, it asks you to choose, with the options laid out, instead of guessing. A mistake in a plan costs one comment.
- Comment on rendered markdown, the editor or a diff
- The agent answers when it is idle, never mid-turn
- Threads move with their lines when the plan changes
Run the whole plan in parallel, checked before it reaches you
A lead agent turns your plan into a task graph: what depends on what, and which agent and model each task gets. You approve it, then SimpleEdit runs it. Independent tasks run in parallel, each worker in a fresh session, and each task that writes code in its own checkout. A task can fan out into one worker per item an earlier task found. Put a task in a loop and a fresh reviewer checks every round until the work is clean. Let Codex review Claude's work, or the other way round.
- Sign-off points open a review of what's built so far; ask for changes and the lead reworks them
- Have the lead change the running plan, and see what changed since you approved
- Step into any worker's terminal and steer it
- The lead wakes only when something needs a decision
The agent that wrote it tells you where to look
When the work is ready, the agent that wrote it asks you for a review. It brings a briefing: what changed, why, the impact, a reading order, and what it ran to check. It also pins its own doubts to the lines in question, so you start where it was least sure.
- A reading order from the agent that knows the change
- The author's questions arrive as threads on the diff
- Its evidence is shown as "Author ran", never as verified
One reviewer per angle, one short list
Press Deep review and SimpleEdit sends one reviewer each for soundness, intent and tests, plus types and architecture if you turn them on. A final pass dedups, ranks and prunes what they found into one short list, so you read a handful of findings instead of a pile of comments.
- Up to five reviewers; soundness, intent and tests run by default
- Tune each lens, or rewrite its prompt to match how your team reviews
- The same review for your agents' PRs and your colleagues'
A sparring partner, not a bot
In the Review Room, a second agent that didn't write the change works through it with you. Ask the Companion whether the author's worry is real. It reads the code at the PR head and answers in the thread, and it can't edit or post anything. Once you know what you want, send the thread to the agent that wrote the change and answer on the line. It replies and fixes the code.
- Send any thread to the Companion, to the author, or into your GitHub review
- A stack of PRs opens as one review, layer by layer
- Your own agent gets your review directly; GitHub gets only what you wrote or accepted
Not just a terminal in a window
Every session gets SimpleEdit's tools, so your agent can use the app, not only print into it. It shows you a diff instead of pasting one. It asks you to choose, with the options laid out. It parks a follow-up on the backlog, messages another agent, or requests your review. The conversation moves out of the scrollback and onto the work.
- Show: diffs, panels, choices to pick from, the right branch
- Hand off: start or hand off sessions, park work on the backlog
- Coordinate: message other agents, build and run workflows
- Review: request your review, read your verdict, answer your threads
Your colleagues' PRs, in the same room
Screen PRs reads the recent PRs that request your review, within the scope you set, and sorts them: needs your attention, quick pass, waiting on author. Open a colleague's PR or stack with the same overview, lenses and Companion. Your review goes out on GitHub in your name, and nobody else installs anything.
- Screen on a schedule; a sidebar badge shows what needs your attention
- Pick up where you left off with "Continue reviewing"
- Open any PR, or a pasted link, as a stack
When an agent has a question, it reaches you
Your agents keep running on your computer. Your phone gets a push notification when one needs you. Answer it, sign off so a run can go on, or work through a review with the Companion. Threads and drafts are the same ones you see on the desktop.
- A push notification when an agent or a run is waiting for you
- The full Review Room: threads, Companion and drafts
- Reaches your computer over Tailscale or OpenTunnel
Your agent, unmodified
Claude Code, Codex and OpenCode run in their own terminal UI, with your login, subscription, settings, MCP servers and slash commands. Around them, SimpleEdit adds:
Sessions
Durable agent sessions that survive restarts, each with its own editor, diff and git log. Group, fork, or hand off to a fresh context.
Backlog
Park follow-up work as a prepared session, and start it when you're ready. Your agent can add to it too.
Claude memory, in view
See what Claude Code remembers about your project, with broken links and unindexed files flagged. Fix it like any other file.
A real editor
The Monaco editor from VS Code, with language servers for TypeScript, CSS, JSON, Rust, Python and Go.
Providers and models
Claude Code, Codex and OpenCode, plus local models through Ollama. Pick the model per session.
Usage
Tokens per session and per workflow task, with a cost estimate for Claude.
Try SimpleEdit
Alpha software: expect rough edges. We build SimpleEdit with SimpleEdit, so the sharpest edges get fixed fast.
Free to use · no telemetry
Install on macOS with Homebrew · Windows, Linux →
brew trust pago/simpleedit
brew install --cask pago/simpleedit/simpleedit