AllCrew is a local-first desktop control plane (macOS and Windows) that turns AI-assisted software development into a managed process: one shared board joins tasks, AI agents, project memory, quality checks and human decisions. It is built on a Rust core with a Tauri 2 shell and keeps all code and data on the company's own infrastructure.

Which AI coding agents does AllCrew support?

AllCrew supports Claude Code and Codex out of the box, and any CLI-based coding agent can be added as a target — new targets such as Gemini or OpenCode render on the board dynamically. Every run is consent-gated, sandboxed and works in its own git branch.

Does AllCrew send code or data to the cloud?

No. AllCrew is local-first: the board, run history, project memory and source code stay on the company's own machines, teammates connect over the company's own Tailscale mesh, and external AI providers are used only under the organization's rules. The demo profile ships with zero telemetry uploads, verified by network capture.

What is the AllCrew Board API?

The Board API is a secured HTTP and realtime-stream interface that gives any connected AI assistant a governed view of the whole system of work: tasks, runs, logs, artifacts, project memory and decisions — with permissioned read and write access. Claude Code ships with the Board API description built in, so one integration replaces many.

Do operators need to know Git to use AllCrew?

No. AllCrew explains every state in plain-language outcomes and places a safe repair action beside every failure. Changes are reviewed as outcomes (what changes, what is preserved, what is touched) and merge through a normal fast-forward path; anything can be stopped, restarted or rolled back with state preserved.

AI work, without the chaos. One shared workflow for every agent.

Run tasks across AI agents from one shared board. Keep context, progress, checks and decisions together — on your infrastructure.

Get AllCrew See how it works ↓

Rust core Tauri 2 macOS + Windows Local-first

The problem

More AI shouldn’t mean more management.

Every agent works in its own session. Context gets lost, decisions get forgotten, checks are inconsistent, and people end up keeping everything in sync by hand.

You add AI to save time. Then spend time managing it.

The answer

One place to run AI work.

Tasks, agents, context, checks and decisions stay in one workflow. Local-first: your board, run history and code stay on your infrastructure. Agents get the context they need and keep work moving without starting from scratch.

Board 12 / 46 ⌗ 100.84.29.92:3200 Connect Claude Code + New task Autopilot Live logs

Backlog 4

STF-1204storefront · main

macOSClaude Sonnet 4.6Loop 1

Fix checkout rounding in cart totals #auto-fix

cart.tstotals.tscheckout.spec.ts

All tests pass with exit code 0 from the runner
No production routes touched

▸ Run

STF-1209storefront · main

macOSCodex

Export weekly reports to CSV

reports.tsexport.ts

In Progress 2

STF-1198storefront · main

macOSClaude Sonnet 4.6Loop 2

Migrate auth to passkeys (no session breakage)

auth.tspasskeys.tssession.ts

Existing sessions survive the migration
WebAuthn flow covered by e2e tests

● claude-code · 12 min

STF-1201storefront · main

macOSCodex

Speed up image pipeline

● codex · 4 min

Human Review 2

STF-1187storefront · main

macOSClaude Sonnet 4.6

New onboarding email flow

emails.tsonboarding.tsx

Diff summarized as outcomes — approve or send back

Review

STF-1180storefront · main

Dark-mode settings page

merged to main

run#214 · cloned repo → branch crew/passkeys → tests green → drafting the migration… ▌

How it works

From task to checked result.

Plan

Define the outcome, then break the work into clear steps, rules, checks and approval points. A good process becomes easy to repeat.

Turn good workflows into playbooks

Execute

AllCrew sends work to available agents and machines. Claude Code, Codex or any CLI agent gets the right branch, the right context and a clear task.

Send work to the right agent

Verify

A task is only done when it passes the checks you set. QA, completion rules and risk checks are part of the workflow from the start.

Check the work before it ships

Review

When a decision is needed, people see what changed, what stays the same and what may be affected. Approve it or send it back with a comment.

Bring people in when needed

Iterate

If the work needs another pass, the loop continues. If an agent or model fails, it picks up from the last checkpoint instead of starting over.

Continue from where it stopped

Inside the app

See the work without opening terminals.

Git Tree turns branches, merges and CI into a visual view the whole team can understand. Memory Center keeps project decisions connected and brings the right context into every agent brief.

Git Tree mail-printing-pipeline ⌄ main ● Lock free

Pre-flight→ Rebaser→ CI gate→ Auto-merge→ Branch GC

0 in progress0 needs rebase0 CI running1 human review0 hot files

ALt-20not in maindocs: map multi-brief global search architecture

ALmainlivechore: ignore the ccc code index

АЛmain[task_5b805f0c07] Multi-brief mode with language-agnostic geo discovery

ALmain⎇ Merge 'feature/task_0bc5e95bd6-attempt-1' into integration/2da64368

ALmainML-12: parametrize the operator language, de-hardcode RU→ES

YMmain[task_7b715c6635] ML-9 · Expose peer_language in review payloads #auto-fix

ALmainML-3 · GET /api/languages endpoint

Git Tree — branches, commits and pipeline gates in one view

Memory Center mail-printing-pipeline

Knowledge graph6 nodes · 5 links

Type

Decision ⌄

Title

Windows release signing

Knowledge

What future agents need to know… Save memory

Memory Center — project decisions connected to every agent brief

What's inside

The system around your AI tools.

01

Shared memory

Tasks, runs, changes and project knowledge stay connected, so a new agent can continue from decisions already made.

Context stays

02

Reusable playbooks

Turn proven ways of working into playbooks with clear steps, rules and checks. Once a process works, the whole team can reuse it.

Repeatable

03

Agent orchestration

Spread work across agents and machines based on availability and capability. Run more tasks in parallel without adding more manual coordination.

Coordinated

04

Built-in checks

QA, completion rules, reviews, risk checks and approvals all happen inside the workflow.

Verified

05

Safe recovery

Checkpoints save progress, so an agent or model failure does not reset the task.

Resilient

06

Clear visibility

See what is running, what is blocked, what passed checks and what happens next — without reading logs or asking for updates.

Visible

07

Local-first control

Your board data and run history stay on your infrastructure. You decide who gets access and how external AI providers are used.

Controlled

Board API

Give any AI agent access to the work.

Without AllCrew, each AI sees only its own session

Codex knows its session. Claude knows its own. Neither sees the full workflow, so people have to move context and coordinate between them.

statuses copied by hand
logs checked one window at a time
context moved between tools by people
AI findings never become actions

AI connected to AllCrew

The board is available through a secured API and real-time streams, within the permissions you set.

reads tasks, dependencies and statuses
analyzes runs and failures
reads project memory and the knowledge graph
creates tasks, starts work and retries runs
asks for human decisions

What’s blocking the release? Which tasks keep coming back? What needs a human decision? What can we run next?

With the Board API, AI can read the work, understand what is happening and act on it. One API instead of many integrations: connect an AI once and let it work with tasks, status, project memory and decisions.

What changes

One system. Clear value for every leader.

CTOGive AI-assisted development shared context, clear checks and one managed workflow.

CFOIncrease output without growing headcount and rework at the same pace.

CIOBring separate AI tools into one layer that works with your existing infrastructure.

CISOKeep working data local, control access and define where human approval is required.

COOReplace manual handoffs and waiting with a clear, repeatable process.

CEOTurn AI from scattered experiments into a repeatable way to ship faster.

One board. Shared context. Any AI.

Without AllCrew, every new agent adds more coordination. With AllCrew, AI work stays shared, repeatable and under control. Early access is open.

Get AllCrew inbox@allcrew.app