A graph of your code, in 17 languages
A sync parses files with tree-sitter into functions, structs, traits, enums, imports, impl blocks and call edges. Ask for the impact of a file, a call graph, references or architecture before an agent edits.
For developers and tech leads
Project Orchestrator parses your repo into a graph, turns a plan into waves of independent tasks and hands each agent the notes and decisions it needs. You reach it through MCP, REST or WebSocket.
In the product todayCovered today
Proofs last checked on 7 Oct 2026
“Two agents edited the same file.”
Parallel sessions need an order: which tickets are independent, which touch the same files, which wait. Without it you serialise by hand or you rebase.
“I cannot tell what this change breaks before the agent does it.”
The agent greps. It has no call graph and no list of what depends on the file it is about to rewrite, so you find out in review.
“Every new session, I explain the repo again.”
Conventions and gotchas live in a CLAUDE.md or in your head. What one session learned is gone for the next, and nobody keeps the reason behind a choice.
“I am the scheduler.”
You keep the ticket list, start each session, check it finished, start the next one. The orchestration is your afternoon.
A sync parses files with tree-sitter into functions, structs, traits, enums, imports, impl blocks and call edges. Ask for the impact of a file, a call graph, references or architecture before an agent edits.
Tasks carry depends_on and affected_files. The plan is cut into levels of independent tasks; tasks of one level that declare the same file are pushed into separate waves. The runner starts a wave’s tasks together, four at a time by default.
33 MCP tools with an action parameter, a stdio proxy for your client, a REST API with 478 route registrations and four WebSocket streams for events, chat sessions, plan runs and live activity.
Notes (guideline, gotcha, pattern…) and decisions attach to code entities. Retrieval runs vector search, then spreads activation over weighted synapses between notes. Each task prompt is built with related notes and decisions.
A protocol is a state machine with states, guarded transitions and runs you can simulate. A lifecycle hook reacts to a task, plan or step changing status: cascade to children, call an MCP tool, create a note, raise an alert, start a protocol.
Claude Code is built in and stays the default. You can register OpenAI-compatible, Anthropic, Codex and ACP provider instances, allow them per project, move a conversation from one engine to another, and let Auto pick the model of a task or turn without ever replacing one you chose. Codex is probed for real and uses the machine’s own login; a full turn over Codex or opencode is still unmeasured, which is why they stay alpha.
A native model can start a sub-agent: it gets its own conversation, opened from its card, keeps the parent’s workspace, is never left running, and announces itself when it ends. A Claude Code sub-agent gets a read-only mirror of the same tree, background ones included. At most four children work at once per parent, and deleting a parent deletes its children.
Swap an HTTP client in one service. Same repo, same Claude Code, without and with Project Orchestrator.
YouYou split the work into tickets in a doc and open three agent sessions by hand.
AssistantTwo sessions rewrite the same client module. The second overwrites half of the first.
AssistantA fresh session proposes the retry policy you rejected last week. It never saw that thread.
YouYou play scheduler: watch each session, start the next ticket, rebase by hand.
AssistantThrough the MCP tools it creates a plan: tasks with depends_on and affected_files, and steps with a verification each.
Project Orchestratorplan get_waves returns the levels. Two tasks that declare the same file land in different waves, and the conflict is listed.
YouYou read the waves, fix a dependency, then start the run with the plan id, the working directory and the project slug.
Project OrchestratorWave 1 tasks start together. Each prompt carries related notes and decisions, so the rejected retry policy is in front of the agent.
Project OrchestratorAfter each wave PO collects agent worktree commits, runs the build check, then starts the next wave. You follow it on /ws/run/{run_id}.
You review one plan and its run instead of scheduling sessions. Quality still depends on your model, your tests and your review.
You say “ticket”, “agent”, “RFC”. Project Orchestrator says Task, Assistant, Proposal. Jargon is welcome on this page.
“a ticket, a unit of work for an agent”
Task
A unit of work inside a plan.
“the ticket breakdown, the task graph”
Plan
How an objective gets done: ordered tasks and steps.
“a level of the dependency graph, a parallel batch”
Wave
A group of tasks that no longer wait for each other: they can start together, the next wave starts when this one is done.
“agents, sub-agents”
Assistants
A conversation with an AI that works on your project; Today shows which ones are running.
“the code graph, impact analysis”
Code
Files, symbols and how they relate: what the assistant reads before it changes anything.
“MCP tools”
MCP server
External tool servers the assistant can call, and the way other tools talk to PO.
“a state machine for a workflow”
Protocol
A repeatable procedure the assistant follows step by step.
“knowledge clusters”
Skill
A set of knowledge the assistant activates on its own.
“an ADR”
Decision
A choice that was made, with its rationale and alternatives.
Words this page leaves out : seamless, supercharge, 10x, autonomous engineer
SupportedLanguage has 17 variants and Cargo.toml 17 tree-sitter grammars. JavaScript files go through the TypeScript grammar. The README states fewer: the code is the reference. Call edges carry a confidence: 0.90 when resolved through an import, 0.85 in the same file, 0.30 to 0.50 for a global guess.
all_tools() returns 33 tools and a test pins that number; the dispatch table MEGA_TOOL_ACTIONS lists 426 actions, the sum of the action enums (recounted on 2026-11-01; the two last tools added are provider and day_plan). The README and several guides still say 22, and a header comment says 30: they are out of date.
compute_waves groups tasks by dependency level, fails on a cycle, then splits tasks of one level that share an affected file by greedy colouring. Called read-only on a local server on 2026-10-07, it returned waves, edges, conflicts and a summary with max_parallel and critical_path_length.
max_parallel_tasks defaults to 4, a run aborts when its cost ceiling is exceeded, and the runner drives one run at a time. At boot it resumes at most one interrupted run, in the directory it started in. The file has 135 test functions; this review did not run a plan end to end.
routes.rs registers 426 routes, among them /api/plans/{plan_id}/waves, /run, /run/cancel and /run/budget, /api/code/impact and /api/code/search, and the WebSocket endpoints /ws/events, /ws/chat/{session_id} and /ws/run/{run_id}.
The workflow builds the release binaries, runs cargo audit and cargo deny, checks formatting, runs clippy with -D warnings, then library, parser and integration tests. It proves the code builds and passes its tests, not that a plan run works on your repo.
Nothing here is hidden: these are the things you might assume and that are not true yet.
The figures come from reading the code, not from a benchmark. The plan runner was not run end to end for this review.
Conflict splitting uses the affected_files a task declares. A task that edits a file it did not declare can still collide.
The runner does not create git worktrees: agents do. It collects their commits after each wave and reports conflicts.
The runner drives one plan run at a time per server.
The engines beyond Claude Code are not all measured against the real services: Codex is probed for real, but a full turn over Codex or opencode is still unmeasured, and a Claude Code on another machine does not get PO tools in its session.
The /api/hooks/activate endpoint exists for Claude Code hooks, but no command installs that hook for you: you wire it in your Claude Code settings.
Call edges are heuristic. A global guess carries a confidence of 0.30 to 0.50, so treat impact analysis as a lead, not a proof.
No roles or per-user permissions on plans and notes. Team access control is planned.
brew install this-rs/tap/project-orchestrator, or the install script. It puts orchestrator, orch and mcp_server on your PATH. Neo4j and Meilisearch still have to run: docker compose up -d neo4j meilisearch from a clone of the repo.
Open the quickstartorchestrator serve listens on port 8080. Then orchestrator setup-claude registers the MCP server in Claude Code.
Ask your assistant: project create with a name and root_path, then project sync with the slug. Or run orchestrator sync --path . Then try code analyze_impact on a file you know.
Create a plan with tasks that carry depends_on and affected_files, then call plan get_waves before you run anything. Start it with plan run (plan_id, cwd, project_slug) and follow /ws/run/{run_id}.
Read the docsAgent runtime, provider resolution, chat transport and spawn envelope are drawn from the verified diagrams of the repos.
Open the internalsorchestrator setup-claude configures Claude Code only: it runs claude mcp add or edits ~/.claude/mcp.json, and pre-approves mcp__project-orchestrator__*. The mcp_server binary is a stdio proxy that needs PO_SERVER_URL, so any client that can launch a stdio MCP command can try it. The Cursor and OpenAI guides in the repo predate that setup and are not verified.
Project Orchestrator drives a Claude Code you installed; it does not replace it. It adds state that outlives a session (plans with dependencies, notes, decisions), a code graph, and a runner that computes waves. If your work fits in one session and your CLAUDE.md is enough, you may not need it.
The graph and the search index run on your machine (Neo4j and Meilisearch, started by docker compose). What an agent reads is sent to its model provider, as with any agent: Claude through Claude Code, or a provider instance you registered.
Without an auth section in config.yaml every request passes as anonymous, which suits a local setup. With it, protected routes require a JWT, and WebSockets use a one-time ticket from POST /auth/ws-ticket. The remote /mcp endpoint is off unless you enable it.
A run takes max_cost_usd, aborts as budget_exceeded when it is crossed, and runs at most four tasks at once by default. The default ceiling is set in RunnerConfig (src/runner/models.rs): read it before an unattended run. Model usage is billed by your provider.
Seventeen are parsed into symbols: Rust, TypeScript and JavaScript, Python, Go, Java, C, C++, Ruby, PHP, Kotlin, Swift, Bash, C#, Scala, Zig, HCL (.tf) and Dart. Other file types have no symbols in the graph.
Yes. The REST API is the same one the MCP tools call, and the orch CLI wraps plan, task, decision, sync and context. Example: curl localhost:8080/api/plans/<plan_id>/waves.
Download
One-click install
There is no Project Orchestrator account to create.
Free · MIT license · Bring your own AISee what it costs
Version 0.0.19 · 13 Nov 2026
The app installs on a computer. Pick the one you use.
Prefer a terminal?
Homebrew · Mac and Linux. Installs orchestrator, orch and mcp_server.
brew install this-rs/tap/project-orchestratorDocker Desktop
Runs the database and the search engine the app relies on. The app starts them for you. Get Docker Desktop
Claude Code
The default engine for chat. The AI itself is not included: with Claude Code, you use your own Claude account, and the app can install it for you. Other engines (Codex, an OpenAI-compatible model, an ACP agent) are available from Settings — newer, in alpha. Get Claude Code
At first launch
A setup wizard in four short parts checks Docker, installs Claude Code if it is missing and downloads the search model.
Runs the two services that hold the project memory (Neo4j and Meilisearch).
Open the downloaded .dmg, drag Docker into Applications, then open Docker once and wait for the whale icon to stop moving.
Or in a terminal
brew install --cask dockerCheck that it works
Each line should print a version number. If you see « command not found », open a new terminal window and try again.
docker --versiondocker compose versionDocker must be running (open it once) before you start Project Orchestrator. Installation page of the publisher
The assistant Project Orchestrator works with. It uses your own Claude account.
Paste the command in a terminal. It installs everything it needs and asks for your password only if required.
Command
curl -fsSL https://claude.ai/install.sh | bashCheck that it works
Each line should print a version number. If you see « command not found », open a new terminal window and try again.
claude --versionRun claude once to sign in. Installation page of the publisher
Free · MIT license · Bring your own AI See what it costs
Try it: start a task in any window
Download
Pick the file for your system above. There is no account to create.
Open
Open the file. A setup wizard checks Docker, installs Claude Code if it is missing and downloads the search model.
Connect your AI
Chat runs on Claude Code by default; other engines are available from Settings (alpha). If it asks you to sign in, type this once in a terminal.
claudemacOS, Apple Silicon
The app for Macs with an M1 chip or later.
Project.Orchestrator_0.0.19_aarch64.dmg · 87.1 MB
macOS, Intel
The app for Macs with an Intel chip.
Project.Orchestrator_0.0.19_x64.dmg · 89.8 MB
macOS, Intel (macOS 10.15 to 13.3)
For Intel Macs on macOS 10.15 to 13.3. Without local embeddings: use an HTTP provider such as Ollama.
Project.Orchestrator_0.0.19_x64-legacy.dmg · 47.7 MB
Linux, AppImage
One file that runs on most Linux distributions (x86_64).
Project.Orchestrator_0.0.19_amd64.AppImage · 163 MB
Debian and Ubuntu, x86_64
Package to install with dpkg or apt.
Project.Orchestrator_0.0.19_amd64.deb · 98.1 MB
Debian and Ubuntu, arm64
Package to install with dpkg or apt, for ARM64 machines.
Project.Orchestrator_0.0.19_arm64.deb · 94.4 MB
Fedora and RHEL, x86_64
Package to install with dnf or rpm.
project-orchestrator-0.0.19-1.x86_64.rpm · 61.5 MB
Fedora and RHEL, arm64
Package to install with dnf or rpm, for ARM64 machines.
project-orchestrator-0.0.19-1.aarch64.rpm · 56.7 MB
Homebrew
Mac and Linux. Installs orchestrator, orch and mcp_server.
brew install this-rs/tap/project-orchestratorInstall script
Mac and Linux. Latest release into ~/.local/bin. Add --no-frontend for the light variant.
curl -fsSL https://raw.githubusercontent.com/this-rs/project-orchestrator/main/install.sh | shPowerShell
Windows. Installs into %LOCALAPPDATA%\orchestrator\bin.
irm https://raw.githubusercontent.com/this-rs/project-orchestrator/main/install.ps1 | iexDocker
Image for amd64 and arm64, with the web interface.
docker pull ghcr.io/this-rs/project-orchestrator:latest# API only, no dashboard: a version tag with the -api suffix# docker pull ghcr.io/this-rs/project-orchestrator:0.0.19-apiorchestrator, orch and mcp_server, with the web interface included. Lighter versions without it are on the release page.
macOS, Apple Silicon
Programs without the app window: orchestrator, orch, mcp_server
orchestrator-full-0.0.19-macos-arm64.tar.gz · 81.9 MB
macOS, Intel
Programs without the app window: orchestrator, orch, mcp_server
orchestrator-full-0.0.19-macos-x86_64.tar.gz · 82.8 MB
Linux, x86_64
Programs without the app window: orchestrator, orch, mcp_server
orchestrator-full-0.0.19-linux-x86_64.tar.gz · 91.1 MB
Linux, arm64
Programs without the app window: orchestrator, orch, mcp_server
orchestrator-full-0.0.19-linux-arm64.tar.gz · 85.6 MB
Windows, x86_64
Programs without the app window: orchestrator.exe, orch.exe, mcp_server.exe
orchestrator-full-0.0.19-windows-x86_64.zip · 92.8 MB
The desktop app starts Neo4j and Meilisearch in Docker for you. With the command line, one compose command does it: follow the quickstart.
More for developersWhat you should see. A list of checks. On the very first launch the Docker lines say “Setup pending”.
What to do. Wait for the setup to open: the start screen only checks what is installed.
What you should see. A green line says “Docker Desktop is running” under two cards, “Docker (recommended)” selected.
What to do. Keep “Docker (recommended)”. The app looks for Docker Desktop. When it says “Docker Desktop is running”, press Next. If it says Docker is missing or not running, use the button next to the message, then wait: the app checks again by itself. The services use the local ports 7474, 7687, 7700 and 4222, and the app itself uses 6600.
What you should see. Three choices (No authentication, Password, OIDC / OAuth), the first one selected.
What to do. On a computer only you use, keep “No authentication” and press Next. On a shared or public network, choose “Password” and set an email and a password: without sign-in, the app’s server accepts connections from your whole network.
What you should see. Claude Code CLI shows a green check once installed (until then: “Claude Code CLI is required” and an install button), and the search model is marked ready.
What to do. If “Claude Code CLI is required” is shown, press “Install Claude Code CLI”. Then let the app download the search model when it offers to; the size is shown on screen. Next unlocks when both are ready. If you have not signed in to Claude yet, see “If it does not work” below.
What you should see. A green “Configuration saved!” message and a “Restart Application” button.
What to do. Press “Generate Config & Save”. When “Configuration saved!” appears, press “Restart Application”. A switch on this screen controls whether the app checks for updates at start-up.
What you should see. A checklist (Docker Desktop, Neo4j, Meilisearch, NATS, Backend server) turning green one line at a time.
What to do. On this second start the app downloads the three images it runs on, then starts them. Later starts reuse the images and are quicker. When every line is green, the main window opens. If a line says “Taking longer than expected”, wait. If a “Continue anyway” button appears, see “If it does not work”.
macOS 11 or later on Apple Silicon (M1 and later). On Intel: macOS 13.4 or later, or macOS 10.15 to 13.3 with the older-macOS file, which has no local embeddings.
Download Docker Desktop from Docker’s page, open the file, drag Docker to Applications, then open it and wait until it says it is running.
Get Docker DesktopDocker Desktop is open and reports that it is running. It is Docker’s own screen, so it is not reproduced here.
Open the .dmg, drag Project Orchestrator to Applications, then open it from there. If macOS says it cannot check the app, we could not verify how the build is signed: allow it in System Settings, Privacy & Security, with “Open Anyway”.
The three problems you are most likely to meet, with what the app shows.
What you see. The app says “Docker Desktop is not running” or “Docker Desktop is required”.
What to do.
What you see. Neo4j or Meilisearch stays “Unhealthy”, or the app says it failed to start services. The services need the local ports 7474, 7687, 7700 and 4222.
What to do.
Only if you are comfortable with a terminal, this shows which program uses the port:
lsof -i :7687What you see. The setup says “Claude Code CLI is required”, or the chat says it has no Claude login.
What to do.
Type this and press Enter:
claudeWindows, 64-bit.
Download Docker Desktop from Docker’s page, run the installer, then start Docker Desktop from the Start menu and wait until it says it is running.
Get Docker DesktopDocker Desktop is open and reports that it is running. It is Docker’s own screen, so it is not reproduced here.
Run the installer and follow the wizard, then start Project Orchestrator from the Start menu. Windows may show “Windows protected your PC” because no signing certificate is set in the build configuration: choose “More info”, then “Run anyway” if you trust the download.
The three problems you are most likely to meet, with what the app shows.
What you see. The app says “Docker Desktop is not running” or “Docker Desktop is required”.
What to do.
What you see. Neo4j or Meilisearch stays “Unhealthy”, or the app says it failed to start services. The services need the local ports 7474, 7687, 7700 and 4222.
What to do.
Only if you are comfortable with a terminal, this shows which program uses the port:
netstat -ano | findstr :7687What you see. The setup says “Claude Code CLI is required”, or the chat says it has no Claude login.
What to do.
Type this and press Enter:
claudeAppImage for Intel/AMD; .deb and .rpm packages for Intel/AMD and ARM64.
Follow Docker’s page for your distribution, then start Docker and wait until it says it is running. The app also recognises Docker installed without the Desktop application.
Get Docker DesktopDocker is running. It is Docker’s own screen, so it is not reproduced here.
For the AppImage: in your file manager, open the file’s properties and allow it to run as a program, then double-click it. For a .deb or .rpm: open it with your software installer, then start Project Orchestrator from your applications menu.
The three problems you are most likely to meet, with what the app shows.
What you see. The app says “Docker Desktop is not running” or “Docker Desktop is required”.
What to do.
What you see. Neo4j or Meilisearch stays “Unhealthy”, or the app says it failed to start services. The services need the local ports 7474, 7687, 7700 and 4222.
What to do.
Only if you are comfortable with a terminal, this shows which program uses the port:
lsof -i :7687What you see. The setup says “Claude Code CLI is required”, or the chat says it has no Claude login.
What to do.
Type this and press Enter:
claude