briar
Open web app
Open web app
Agent Development Environment

From issue to PR.
Operate your agent work.

Briar is a cloud-coordinated, local-execution Agent Development Environment that connects and observes agent work while running code tasks in repositories on your machine.

Download for MacWeb App
All download options
Supports
  • Claude
  • Codex
  • Cursor
  • Grok
  • Antigravity
  • OpenCode
  • OpenRouter
See how it works
Why Briar

What Briar adds to the agents you already use.

Codex and Claude Code already write code. Briar adds a queue, a record, and a place for a human to approve that work.

  1. 01

    Explicit completion criteria

    Issues move through the stages you define — analysis, implementation, local QA, review, release — so a run finishes because it met its criteria, not because the agent stopped talking.

  2. 02

    An observable run timeline

    Each stage records its evidence — the command that ran and whether it passed — so you can check the work instead of re-reading the conversation.

  3. 03

    Retry-safe event history

    Status updates and evidence are written so a paused or interrupted run resumes from its exact checkpoint and attempt, not from scratch.

  4. 04

    Human approval gates

    You choose where a run pauses for review. It continues only after you approve it.

  5. 05

    A repo-local execution model

    Agents run in an isolated Git worktree on your machine. Only task state and Git metadata sync to Briar — not your source.

Free and open source. Briar is released under the Apache-2.0 license on GitHub. You'll need to sign in to connect a repository.

01 / PRODUCT

Turn agent development
into an operable system.

Move beyond one-off automation to an execution flow that real product teams can trust.

01

Keep code local

Repository source never leaves your device. Briar securely syncs only the task state and Git metadata it needs.

02
CCH

People and agents, together

Connect Codex, Claude, and team approvals in one timeline and workflow.

03

Close the loop

Make completion criteria explicit from context gathering and implementation through QA, review, and deployment.

02 / WORKFLOW

Keep work moving forward.

Turn conversations and issues into executable tasks, and never lose sight of their progress.

  1. 01

    Intake

    Collect issues, feedback, and errors in one actionable task queue.

  2. 02

    Run

    Codex and Claude read context and start work in the real repository.

  3. 03

    Observe

    See every status and decision from analysis through implementation and QA.

  4. 04

    Ship

    Connect verification results and PRs, then deploy only work that meets its completion criteria.

03 / AGENTS

Keep the agents you prefer.

No need to change tools. Briar connects different agents through one operating model.

Active collaborators All systems operational
C
CodexImplementation & local QA
3 running
C
ClaudeReview & reasoning
1 reviewing
H
HumanDirection & approval
2 approvals
04 / SECURITY

Keep code local.
Make trust the default.

Agents run only inside the connected Git root. Source code stays local, tokens are stored as hashes, and risky actions can require human approval.

✓ Repository path never leaves your device✓ Permission-aware agent execution✓ Auditable, retry-safe event history
Ready when you are

Delegate the work.
Trust the result.

Start a development flow where people and agents work together with Briar.

Open Briar on the web Download Briar for Mac Download for Android View on GitHub
Latest release · macOS Apple Silicon · Android companion