Orchestrator MCP

one agent. second opinion. same terminal.

Make Claude Code ask Codex. Make Codex ask Claude Code.
Use the subscriptions already signed in on your computer.
No provider API keys to configure.

See it · Install · What you get · Reviews · Workflow · Security · Dashboard

Orchestrator MCP is a local Model Context Protocol server that lets one coding agent consult another. It launches the Codex, Claude Code, OpenCode, or experimental Antigravity CLI already installed and authenticated on your machine, routes the request, and returns a structured answer.

It does not ask for a provider key, proxy provider traffic, or silently switch models. Authentication remains inside each vendor's CLI.

Before / After

Same subscriptions. Less context shuffling.

 Claude Code host  ──►  Orchestrator MCP  ──►  Codex CLI
 Codex host        ──►  Orchestrator MCP  ──►  Claude Code CLI
 Any host          ──►  Orchestrator MCP  ──►  OpenCode CLI (DeepSeek, Qwen, local…)
 Any host          ──►  Orchestrator MCP  ──►  Antigravity CLI (experimental)

                         local routing
                    no provider API keys
                   no same-runtime loops

The host's own runtime is always excluded. Claude Code cannot consult Claude Code through this server, and Codex cannot consult Codex.

Install

Homebrew:

brew tap crAK1644/tap
brew install orchestrator-mcp-server

Apple Silicon uses a prebuilt package. Intel macOS and Linux build dependencies from source; use the uvx option if you want a faster, temporary install.

1. Sign in to the agent CLIs

Sign in to each agent you want Orchestrator to use:

codex login
claude auth login

These are the normal Codex and Claude Code login flows. Orchestrator checks readiness, but never reads or stores their credentials.

For OpenCode, sign in once with opencode auth login for whichever provider you plan to consult. Hosted providers only — this server does not run a model on your machine. See the OpenCode runtime section below.

2. Create config.yaml

consult:
  database_path: ~/.orchestrator-mcp/consultations.sqlite3
  timeout_s: 180

  agents:
    codex:
      runtime: codex
      command: codex
      model: gpt-5.6-sol
      priority: 10
      web_search: true
      scores: { coding: 95, research: 90, reasoning: 95, review: 90 }

    claude:
      runtime: claude
      command: claude
      model: claude-opus-4-6
      priority: 10
      web_search: true
      scores: { coding: 90, research: 95, writing: 95, review: 95 }

See config.example.yaml for every option, an OpenCode agent, and an experimental Antigravity example.

3. Add the server to your MCP client

claude mcp add orchestrator \
  --env ORCHESTRATOR_CONFIG=$PWD/config.yaml \
  --env ORCHESTRATOR_HOST_RUNTIME=claude \
  -- orchestrator-mcp-server

[mcp_servers.orchestrator]
command = "orchestrator-mcp-server"
env = { ORCHESTRATOR_CONFIG = "/absolute/path/to/config.yaml", ORCHESTRATOR_HOST_RUNTIME = "codex" }

Restart the MCP client after changing its configuration.

[!TIP] Use an absolute ORCHESTRATOR_CONFIG path. GUI-launched clients often start in a different working directory and inherit a smaller PATH than your terminal.

Run with uvx instead

claude mcp add orchestrator \
  --env ORCHESTRATOR_CONFIG=$PWD/config.yaml \
  --env ORCHESTRATOR_HOST_RUNTIME=claude \
  -- uvx orchestrator-mcp-server

[mcp_servers.orchestrator]
command = "uvx"
args = ["orchestrator-mcp-server"]
env = { ORCHESTRATOR_CONFIG = "/absolute/path/to/config.yaml", ORCHESTRATOR_HOST_RUNTIME = "codex" }

What you get

The consultation tools

These deletion tools remove local SQLite records only. They cannot erase a consulted runtime's own CLI or provider history.

Three independent opt-ins: the consult tools are always advertised, the review tools only with a consult.review block, the workflow tools only with a consult.workflow block. Reviewers are not a workflow, and a workflow is not reviewers.

Slash commands

The server also serves MCP prompts, which a client that speaks prompts/list renders as slash commands. In Claude Code they appear as /mcp__<server-name>__<command>, where the server name is whatever you called it in your MCP client config — /mcp__orchestrator__review for the orchestrator entry shown above.

Every argument is optional; a command with none expands into an instruction to ask you for the missing part. review and workflow are advertised only when their tools are, on the same two answers — a command that could only reply "no reviewers are configured" costs a round trip and reads like a bug.

Two things worth being clear about. Nothing is installed. These arrive over the same stdio connection as the tools: no command directory, no generated markdown, nothing written to your machine, and a client that does not speak prompts/list is unaffected. A prompt is text, not an action. Expanding one consults nobody, sends nothing, and starts no workflow — it reaches the host's conversation as if you had typed it, and the host then calls the tools, checkpoints and all. They exist because the flows worth having here are handshakes, and a host driving them from tool descriptions alone tends to skip the checkpoint that makes them worth having.

How consultation works

orchestrator_consult selects the eligible agent with the highest capability score. Lower priority wins a score tie; agent ID breaks the final tie. A missing capability or a score of 0 makes an agent ineligible.

The selected CLI runs under its existing login and returns one response envelope:

If the chosen agent fails, Orchestrator returns that failure. It does not quietly fall through to a different model.

Choose the evidence source

Reviews, with a checkpoint

A consultation asks one agent. A review asks one or more configured reviewers the same question over the same material.

 plan review          approve + run          synthesize
 sends nothing   ──►  reviewers answer  ──►  host records conclusion
      │                    in parallel                │
      └─ scope              one-time token            └─ every Critical kept
         reviewers
         secret hits
         request count

Enable reviews in config.yaml:

consult:
  review:
    reviewers: [codex]          # standard: exactly one
    deep_reviewers: [codex, claude]  # deep: one to five

The workflow is deliberately split:

1. orchestrator_review creates a plan and sends nothing. The plan shows reviewers, material size, web access, request count, and locations of credential-shaped text.
2. Show that plan to the user. orchestrator_review_run spends its one-time token and asks reviewers in parallel.
3. Read every result and call orchestrator_finalize_review. Reviewer replies alone leave the review at awaiting_synthesis.

The checkpoint binds the scope and makes the token single-use, but it is advisory: MCP gives the server no separate human channel, so it cannot prove who saw the plan. Human approval depends on the calling client's tool-confirmation experience or an external gate.

Finalization must preserve every machine-readable Critical finding, even when other reviewers disagree with it. Deep mode also requires the host agent to record its own findings before seeing the reviewers' answers.

[!IMPORTANT] Material sent to a reviewer may remain in that vendor CLI's own history. Orchestrator cannot erase Codex, Claude Code, OpenCode, or Antigravity session logs.

The three-phase workflow

A consultation is one question. A review is one body of material. A workflow is a whole job held together: research and planning, implementation and testing, review and fixing, the last two in a capped loop, with every consultation and review it produces hanging off one workflow_id.

 research → plan → author_execution_prompt → implement → test → review → synthesize
                                                 ▲                          │
                                                 └────────── fix ◄──────────┘
                                                   capped at max_fix_rounds

Enable it in config.yaml. Without a workflow: block the workflow tools are not advertised at all, the same rule the review tools follow:

consult:
  host:
    runtime: claude              # asserted against ORCHESTRATOR_HOST_RUNTIME
    model: claude-opus-5         # optional; see "Host identity" below
  workflow:
    max_fix_rounds: 5
    roots: [~/src]
    advance_on_failed_test: false
    review_policy:
      different_from_implementer: true
      different_from_planner: false
    bindings:
      research:  {agent: codex-sol}
      plan:      {agent: codex-sol}
      author_execution_prompt: {executor: host}
      implement: {agent: deepseek-flash, execution: patch}
      apply_patch: {executor: host}
      test:      {executor: host}
      review:    {agents: [codex-sol, claude-opus]}
      synthesize: {executor: host}
      fix:       {agent: codex-sol, execution: patch}

workflow: requires store_full_content: true and refuses at startup otherwise. A workflow is its stored plans, briefs, patches and reports, and the review step cannot finalize without them — the failure belongs at boot, not several paid steps in.

Models are not tied to phases

A binding is one of three shapes, and mixing them is a startup error:

Leaving agent: out means auto: routed by capability score for the step, ties broken by priority then agent id, the same rule orchestrator_consult uses. GPT, Claude, DeepSeek, Gemini, Qwen — any model reachable through a supported runtime can take any step it scores for and is permitted the mode for. A step with no binding falls to the host, which is the conservative default — with one exception. review cannot be the host's: its product is a review row that only the review service writes, and a host-recorded outcome would be a synthesis written straight into a step. Because unbound steps default to the host, a config that simply never mentions review is refused at workflow_start, before anything has been spent, rather than at the review step after research, planning and implementation have all been paid for.

Steps route on the four capabilities added for this: planning, prompt_authoring, testing and synthesis, alongside coding, research and review.

Bindings are resolved and snapshotted at workflow creation, and so is the policy they run under — the round cap, advance_on_failed_test and the review policy. Editing config.yaml does not reroute a running workflow or move its cap; that takes orchestrator_workflow_plan_replan and its own approval. A replan re-decides the steps you name and leaves every other one on the routing the workflow already had.

Execution modes, and what each one can reach

execution_modes on an agent is operator trust, not capability. What actually happens is the intersection of that list with what the runtime can be made to do, and a refusal names which side said no.

A root allowlist and a prompt instruction are not containment. An agent that declares isolated_write on a runtime that cannot be contained is refused at startup, not at routing time.

No delegated agent writes to your working tree. In patch mode the agent never sees your checkout — only the context the host sent it, exactly as a reviewer does — and the diff comes back for the host to apply. In isolated_write it sees a copy: a worktree under ~/.orchestrator-mcp/worktrees/<workflow_id>/<step_id>/, checked out at the latest applied result (the workflow's baseline until the host has applied anything) and deleted as soon as the diff is captured. Either way the step ends in awaiting_host_apply and the host owns the branch. ConsultAdapter gained nothing for any of this: it is still three verbs with no way to ask for anything agentic, and write capability lives in a separate package behind a separate protocol.

Four things about a contained run are worth knowing before you use one:

• The diff is the record, not the reply. Orchestrator runs git add -A in the worktree and takes the staged diff against the baseline, so files the agent created are captured too — a plain git diff <baseline>.. would miss them. The model's summary is stored beside the patch as a description of its work, never as the account of it. Its list of commands is stored the same way, and is knowingly incomplete: codex omits sandbox-denied commands from its event stream entirely.
• The agent cannot commit. A worktree's git directory lives outside its sandbox, so git add, commit, stash and checkout all fail from inside. It leaves the work in the tree and this server records it. The step's timeout is consult.workflow.execution_timeout_s (900s by default), not consult.timeout_s, which is sized for a question.
• Two things a diff cannot carry, and neither passes in silence. A repository created inside the worktree — a scaffolded subproject, a vendored fixture, any git init — makes git add -A refuse the whole tree. The step fails and the worktree is kept, with its path in the error, because at that moment it holds the only copy of the work. Ignored files are skipped by git add -A by design and can never appear in a patch; they come back listed on the step's ignored field rather than vanishing with the worktree.
• The repository capture reads is the one pinned before the agent started. A worktree's .git is a one-line file naming its git directory, and it sits in the directory the agent spent the whole step writing to. That gitdir is read at worktree creation and passed explicitly to every later git call, so what capture diffs is not decided by a file the step could rewrite.

Host identity

(runtime, model) is an execution identity, not an agent id, and it comes from trusted startup configuration only — never a tool argument. runtime still comes from ORCHESTRATOR_HOST_RUNTIME, which stays the authority; naming it under host: is an assertion, and a mismatch refuses the boot.

model is the part the environment cannot supply. With it, a different model on the host's own runtime becomes routable. Without it, every agent on that runtime is excluded — today's consult behaviour. Write the versioned name: identity must be provably different, so opus and claude-opus are treated as claude-opus-5 and refused, and a name with no version at all is refused for being unprovable rather than assumed distinct.

review_policy is checked against this resolved identity too, so two agent ids pointing at one model are one reviewer however they are spelled.

Two calls per step

plan_step                    run_step | record_host_step
returns a preview       ──►  spends that step's token
and a one-time token         and runs or records it

There is no one-call form and no workflow-level token. One approval covering research, implementation, every fix round and every review would be an approval of nothing in particular. What a token proves is snapshot integrity — that the step being run is the step that was previewed, with the same agent, mode and inputs. It cannot prove a human saw anything, because MCP gives this server no channel to one; that depends on your client's tool-confirmation experience or an external gate.

A workdir must resolve beneath a configured root. / is never accepted and nothing is inferred from the working directory. A dirty tree is refused without allow_dirty.

A delegated step sees only what the host hands it

Nothing here reads your repository on an agent's behalf, so a delegated step starts with no code in front of it. orchestrator_workflow_plan_step(workflow_id, step, context) takes that material — the source of the files being changed, most of the time — and the host decides what goes in it. Without it, an implementation step has the plan and the brief and nothing to patch, and the honest models say exactly that instead of inventing a file they were never shown.

The material is redacted with the same scrubber as everything else before it is stored and before it is sent, and it is covered by the step's prompt hash, so the text the preview described is the text that goes out. A review step gets the same material its coding steps did.

Tests are observed, not claimed

A coding agent's statement that it ran the tests is retained as reported information; it is never the test result. A TestReport carries the exact command, working directory, exit code, bounded output, duration and the commit tested, plus reported_by, which the service assigns from which tool wrote the row and never reads out of a caller's payload. A host-written report is host-attested, and its provenance says so.

reported_by: orchestrator has exactly one source: a test step bound to isolated_write, where the exit codes come out of the CLI's own event stream rather than out of anything the model wrote. Read what that does and does not claim. It means every command the run reported returned zero — codex omits sandbox-denied commands from that stream, so it is not a claim that the project's suite ran. A contained test step that edited files while testing lists them on the report's changed_files; the worktree is deleted either way, so nothing there is applicable, but "the tests pass" and "the code was edited until they did" no longer look identical.

A failed test returns to fixing rather than advancing, unless advance_on_failed_test is set.

Two fields are cross-checked rather than stored side by side. A report cannot be passed with a non-zero exit code or with none at all — a command whose exit code was never read is skipped, which is what a denied command or a killed process produces — and it cannot be failed with a zero. The commit is stamped by the service from what the workflow currently holds, not taken from the payload: loop_done compares them, so a pass from an earlier round cannot carry a later one.

apply_patch is the one step whose whole job is a side effect on your tree, and the only evidence it happened is a commit that was not there before. Recorded without one, or with the commit it started from, the step is refused and marked failed rather than advancing — there is no applied: false to write, because a step that did not do its work is a failed step. Plan it again once the patch is applied and committed.

Where the loop stops

The review step goes through the review service — it does not write a synthesis straight into workflow storage, which would route around the guarantee that every serious finding survives. Findings gained a disposition (open, fixed, rejected, accepted_risk), because "unresolved finding" was previously not representable; rejecting or accepting the risk of a critical or important finding without a reason is refused.

loop_done is computed here, never asked of a reviewer. It is true only when the authoritative tests passed for the current commit, every reviewer's findings parsed and were retained, no critical or important finding is still open, missing_serious passed, and the workflow is in no exceptional state. Reaching max_fix_rounds with serious findings still open ends the workflow needs_attention — not completed.

A fix round carries the findings that are still open, read back from the review row rather than from workflow storage, so the round is an answer to the review and not a second pass at the goal. A re-review gets them too.

The execution contract, honestly scoped

The coding prompt is assembled by this code: our EXECUTION_CONTRACT first, then the scope, accepted plan, authored brief and prior findings as a JSON payload. An authored brief is data inside that payload, and no field turns it into contract text.

That is code ownership, not a transport-level enforcement boundary. Claude Code has a real system-prompt channel; Codex and OpenCode receive one compiled text, so there the ordering is a prompt convention a determined model could argue with. Saying so is more useful than overclaiming.

Security model

[!WARNING] Redaction covers every retained database copy, and only this database. Credential-shaped values are replaced before every insert while the target CLI still receives the original material. Detection is best-effort pattern matching rather than a scanner with perfect recall, so a secret with no recognizable shape can survive it. Keep the database private, or set store_full_content: false.

Vendor history is outside all of this. Material sent to a reviewer also lands in that reviewer's own CLI history — Codex writes ~/.codex/sessions/, and the others keep their own logs. Orchestrator cannot redact or erase those files. It does read from them, in three places and for two fields: the Codex adapter opens the rollout file for the session it just ran to recover the model identity the CLI does not otherwise report, and opens the newest rollout to read the latest Codex CLI rate-limit numbers; the OpenCode adapter runs opencode export on the session it just ran, for the same reason — the model identity is absent from that runtime's event stream. Nothing else is taken from any of them.

Two more limits worth knowing:

• CLI error text is shortened and common secret formats are redacted, but an unusual one may still appear in a returned error. Do not forward a raw error envelope somewhere untrusted.
• A caller-supplied JSON Schema is trusted input. A pathological regular expression in one can consume a large amount of CPU.

Orchestrator checks structure, routing, permissions, and model identity where observable. It cannot prove that a model's factual claims are true.

    deepseek-flash:
      runtime: opencode
      command: opencode
      model: opencode/deepseek-v4-flash-free
      scores: { coding: 60, reasoning: 60 }

Local dashboard

The optional dashboard shows agents, routing decisions, prompts, answers, usage, latency, errors, reviews, and recorded fix rounds. It is off by default because it can display everything stored in the consultation database.

consult:
  dashboard:
    enabled: true
    editable: false

Start it separately:

ORCHESTRATOR_CONFIG=/absolute/path/to/config.yaml orchestrator-mcp-dashboard

Open http://127.0.0.1:8765.

Set editable: true to manage consult agents and reviewer selection in the browser. Browser-managed agents are written to ~/.orchestrator-mcp/agents.yaml; the dashboard never rewrites config.yaml, runs login commands, or starts consultations.

Both the MCP server and dashboard read configuration at startup. Restart them to pick up changes.

Configuration

ORCHESTRATOR_CONFIG points to the YAML file. If unset, the server looks for config.yaml in its working directory.

consult is the only top-level section. Configuration from releases before 0.4 containing capabilities, model_list, router_settings, or limits is rejected at startup because direct API routing was removed.

System requirements

• macOS or Linux. Windows is not currently tested.
• Python 3.11, 3.12, or 3.13.
• Homebrew or uv.
• A stdio MCP client such as Claude Code or Codex.
• At least one other supported agent CLI installed and signed in.

Test it

The offline suite uses fake CLI agents. It needs no network and spends no model capacity:

uv sync
uv run pytest -q

Live smoke tests use the agents in your configuration:

ORCHESTRATOR_HOST_RUNTIME=claude uv run python smoke_consult_live.py
ORCHESTRATOR_HOST_RUNTIME=claude uv run python smoke_review_live.py

Live tests make real requests and may use paid capacity. Do not run them in CI unless that is intentional.

Troubleshooting

Deliberately not included

• No direct provider API routing or provider API-key configuration.
• No file edits, shell commands, MCP tools, or subagents for consulted agents.
• No automatic fixes; the host agent owns edits and tests. A workflow records and validates the phases, it does not run the job unattended.
• No delegated write to your actual working tree. isolated_write runs in a throwaway worktree and is codex-only; every other runtime refuses it, and the host applies every patch.
• No streaming; each consultation returns one complete envelope.
• No dashboard-initiated consultations.
• No automatic configuration reload.
• No multi-user or shared state.
• No account system for the loopback dashboard.

Contributing

Issues and pull requests are welcome.

1. Fork the repository and create a branch.
2. Make the change and add a test that fails without it.
3. Run uv run pytest -q.
4. Open a pull request.

Keep private configuration, login data, and consultation databases out of commits. For bugs, open an issue with the response envelope after removing paths, credentials, and other private information.

License

MIT · PyPI · GitHub issues

Built with Pydantic and the Python MCP SDK.