Run a fleet of agents.See all of it.

Every agent works in its own worktree with your team’s context, and CodeSense shows you what each one is doing.

Apply for the private beta See what teams do
A full CodeSense window: the context tree and a rendered story doc on the left, the pipeline dependency graph and notes on the right, four live agent terminal sessions side by side below, and workspace tabs along the top and bottom.
Owner's workspace Real CodeSense app Full window capture
  • A full CodeSense window: the context tree and a rendered story doc on the left, the pipeline dependency graph and notes on the right, four live agent terminal sessions side by side below, and workspace tabs along the top and bottom.

    Side by side

    Available now

    Run agents side by side, each in its own worktree terminal, with live diffs. No multiplexer needed.

  • Activity

    Beta Needs an agent that posts to Activity

    The Activity panel lists what you owe your agents and what they owe you, with a count of what needs you and a notification when you’re away from the app.

What teams do with CodeSense

Seven jobs, one app. Every card says how real it is today.

One loop, all in files you own

CodeSense gives every agent you launch your team’s context in its own worktree, across all your repos and whichever agent you run, and turns your stories into agent work in the right repo.

  1. Context Your team’s rules and skills land as files in each agent’s worktree when it launches.
  2. Plan Story files carry status, dependencies and sprint, and the nav shows what’s ready.
  3. Agents Your agent works in its story’s own worktree, already knowing your context.
  4. Done The agent says when it’s done, and you see the result. Status is a field in the story file that you or the agent move; nothing gates the move.
  5. Dashboards Pages you write show the work, live, beside your plans.
  6. Flows back
    Beta Needs an agent that calls reverse_sync_worktree
    An agent can publish a skill improved in its worktree back to the team.

Then the loop starts again at step 1, with what the agents learned.

Why it compounds: the story file is at once the tracker row, the agent’s input and the dashboard’s data.

Write a skill once. Each Claude Code agent you launch in your team’s repos gets it as a real file, hidden from git, as does any agent whose context names it.

How context works
File .claude/skills/review-pr/ SKILL.md Markdown 
---
name: review-pr
description: Review a pull request the way the billing team does.
---

# Review a PR

1. Money is a Money type in bigint cents. Flag any float.
2. Every POST that moves money takes an idempotency key.
3. Run `npm test` before you approve.
The CodeSense sync view for a sample team: four changed files, with a newly added write-migration SKILL.md shown in full and the team's commit history alongside.
Real CodeSense UI · sample project Build a skill once. A new skill in your team context, shown as a change ready to sync through git.

A story’s status is a field in its file. You or your agent can move it, and nothing in CodeSense blocks the move. A status outside your team’s list is kept and shown as drift.

How plans work
File planning/checkout-v2/ ll-214-refund-endpoint-idempotency-keys.md YAML 
---
type: story
story_id: LL-214
title: Refund endpoint idempotency keys
status: review
source_repo: ledgerline/api
depends_on: [LL-209]
---
A story written as HTML, a wallet pay button, rendered in CodeSense with a flow diagram and acceptance-criteria and rollout tables, beside the epic tree with each story's status.
Real CodeSense UI · sample project An HTML story, rendered in the app beside its epic tree.

A dashboard is an HTML file in your context. Its buttons open workspaces, run skills and move stories, with no permission prompts.

Dashboards that do work
File dashboards/ sprint-14.html HTML 
<!doctype html>
<html>
<head>
  <title>Sprint 14</title>
  <meta name="cs:type" content="dashboard">
</head>
A sample Sprint 14 lead-view HTML page rendered in CodeSense, with Blocked, In review and Awaiting me cards, each with action buttons.
Real CodeSense UI · sample project A dashboard page written for the sample sprint, rendered beside your plans.

Own it. Share it.

  • Plain files in your git.

    Your plans are plain files you already know how to edit: markdown with frontmatter, or rich HTML pages. There’s no database and no migration, and any extra key becomes metadata you can query.

    The app runs on your Mac, your context and plans are files in your git, and CodeSense never calls a model itself.

  • Your team's context is a git repo.

    Your team context is an ordinary git repo. You, or an agent you ask, commit and push a change there, and teammates see it as incoming; CodeSense pulls it for them when it fast-forwards cleanly, or they review the merge.

From watching to trusting

Three rungs. One you can stand on today.

  1. Watch

    Available now

    Every agent side by side, each in its own worktree terminal.

  2. Delegate

  3. Trust

Start in an hour

Install, open your repos, and run your first agents with your team's context.