بنقرة واحدة
animus-mcp-setup
Set up .mcp.json, Claude Code permissions, and connect AI tools to Animus's MCP server
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
القائمة
Set up .mcp.json, Claude Code permissions, and connect AI tools to Animus's MCP server
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
Authoring policy for producing Office documents (xlsx/pptx/docx) as `document` subjects. Used by the author-document workflow's generate phase — draft a spec, render+store via the animus-document-engine store_document tool, then record it with create_subject.
Operate Animus v0.7 execution environments — environment plugins and coder nodes, the cross-phase environment broker, workflow/phase `environment:` pinning, `environment_routing:` rules, `workspaces:` multi-repo checkout sets, lease records and the startup reaper, and animus-environment-railway. Use when pinning workflows to remote execution, debugging node acquisition or teardown, or authoring multi-repo runs.
Operate the Animus portal (animus-launchapp) — the flat-named portal MCP surface and admin-only tool visibility, Claude/Codex subscription Connections, the external MCP-server catalog with Connect handshake and ext.* subject projection, the durable script registry and phase_context_schema authoring loop, team_* workflow authoring over MCP, and trigger-event observability. Use when driving a portal deployment rather than a local CLI install.
Product lifecycle agents — product owner, architect, auditor, docs-writer, devops, researcher personas
Guide a project from idea to autonomous engineering setup — interview the user, write VISION.md, AGENT_PRINCIPLES.md, registry, agents/workflows/phases/schedules YAML, scripts, and a first runnable task. Use when standing up Animus in a new project beyond the minimal /animus-setup scaffold.
Build Animus workflow packs - `pack.toml`, workflow exports, runtime overlays, MCP server descriptors, schedules, and marketplace operations. Use when creating or updating installable Animus packs.
استنادا إلى تصنيف SOC المهني
| name | animus-mcp-setup |
| description | Set up .mcp.json, Claude Code permissions, and connect AI tools to Animus's MCP server |
| user_invocable | true |
| auto_invoke | true |
| animus_version | 0.7.0-rc.18 |
Animus exposes its operations as an MCP server. MCP-aware assistants can call
typed tools instead of shelling out to animus.
After installing Animus, verify the binary path:
which animus
animus --version
Create .mcp.json in the project root:
{
"mcpServers": {
"animus": {
"command": "animus",
"args": ["--project-root", ".", "mcp", "serve"]
}
}
}
The relative --project-root . resolves against the assistant's launch
directory. Use an absolute path if the assistant is launched elsewhere.
If the launch environment strips PATH, pin the binary returned by
which animus:
{
"mcpServers": {
"animus": {
"command": "/Users/you/.local/bin/animus",
"args": ["--project-root", "/Users/you/my-project", "mcp", "serve"]
}
}
}
animus mcp serve accepts three flags worth knowing:
--management — also exposes the animus.interactions.* inbox tools
(list/answer pending agent questions and approvals). Off by default so
agent-injected servers can never answer their own approvals. Use it for
your own assistant's .mcp.json if you want to triage the escalation
inbox over MCP (the CLI alternative is animus agent interactions).--agent-id <ID> — pins the identity used by the blocking
animus.agent.ask / animus.agent.request_approval tools (env fallback:
ANIMUS_MCP_AGENT_ID), so a payload agent_id cannot select a looser
approval_policy.--workflow-id <ID> — pins the workflow context and flips the escalation
tools' default wait mode from block to suspend (env fallback:
ANIMUS_MCP_WORKFLOW_ID).The pins are appended automatically on the agent-injection paths
(animus agent run --agent, workflow phases); set them yourself only when
hand-wiring a server for a known agent.
Do not put API keys in .mcp.json env blocks. Store credentials in the OS
keychain with animus secret set <KEY> (preferred, v0.5.8+); workflow YAML
${VAR} interpolation and plugin spawning check keychain entries when the
env var is unset.
Restart the assistant after creating or editing .mcp.json, then test:
animus.daemon.status.animus.subject.list with { "kind": "task", "limit": 5 }.animus.workflow.config.validate.If subject tools fail with a missing backend, the error's remediation
payload carries the exact install command. Typically:
animus plugin install-defaults
animus daemon preflight
(Flagless install-defaults installs the active flavor's full required set —
provider, both subject backends, transport, workflow runner, queue — as of
v0.5.14.)
Exact tool counts drift between releases — enumerate the live surface with
animus.tools.list (grouped catalog) rather than trusting a doc snapshot.
The two animus.interactions.* tools are gated behind --management.
Families and their purposes:
| Prefix | ~Tools | Purpose |
|---|---|---|
animus.agent.* | 12 | Agent runs, profiles, memory, messages, ask/approval escalation |
animus.interactions.* | 2 | List/answer pending agent questions and approvals (mcp serve --management only) |
animus.daemon.* | 12 | Daemon lifecycle, monitoring, and the observe front-door |
animus.cost.* / animus.budget.* | 3 | Breach log (cost.decisions) + fleet budget get/set (budget.get/budget.set) |
animus.subject.* | 8 | Task, requirement, and external subject CRUD plus batch create/update |
animus.queue.* | 7 | Dispatch queue management (bulk subject_ids[] on hold/release/drop) |
animus.workflow.* | 17 | Workflow execution, phases, gate approve/reject, config, checkpoints |
animus.output.* | 6 | Run output, JSONL, monitor, artifacts |
animus.logs.* | 1 | Active log backend tailing |
animus.skill.* | 5 | Skill list, resolve, search, create, update (project or user scope) |
animus.memory.* | 4 | Project-scoped agent memory |
animus.plugin.* | 9 | Plugin control and marketplace tools |
animus.tools.* | 2 | Tool discovery: search (ranked intent search) and list (grouped catalog) |
animus.task.* and animus.requirements.* are gone — use animus.subject.*
with kind: "task" or kind: "requirement". The animus.runner.* tools were
removed in v0.5.13: use animus.daemon.health (provider_plugins_healthy)
for runner/provider health and the CLI's animus doctor --fix for orphan
cleanup.
To auto-approve selected Animus MCP tools, add entries like these to
.claude/settings.local.json. If your MCP server name is animus, Claude
Code tool ids look like mcp__animus__animus_subject_list (dots become
underscores; hyphens inside a verb are kept, e.g.
mcp__animus__animus_daemon_config-set).
{
"permissions": {
"allow": [
"mcp__animus__animus_daemon_health",
"mcp__animus__animus_daemon_status",
"mcp__animus__animus_daemon_start",
"mcp__animus__animus_daemon_stop",
"mcp__animus__animus_daemon_events",
"mcp__animus__animus_daemon_logs",
"mcp__animus__animus_daemon_observe",
"mcp__animus__animus_daemon_agents",
"mcp__animus__animus_daemon_config",
"mcp__animus__animus_subject_list",
"mcp__animus__animus_subject_get",
"mcp__animus__animus_subject_create",
"mcp__animus__animus_subject_update",
"mcp__animus__animus_subject_status",
"mcp__animus__animus_subject_next",
"mcp__animus__animus_queue_list",
"mcp__animus__animus_queue_enqueue",
"mcp__animus__animus_workflow_list",
"mcp__animus__animus_workflow_run",
"mcp__animus__animus_workflow_get",
"mcp__animus__animus_output_tail",
"mcp__animus__animus_output_run",
"mcp__animus__animus_output_phase-outputs",
"mcp__animus__animus_cost_decisions",
"mcp__animus__animus_logs_tail",
"mcp__animus__animus_plugin_list",
"mcp__animus__animus_skill_search",
"mcp__animus__animus_tools_search",
"mcp__animus__animus_tools_list"
]
},
"enableAllProjectMcpServers": true
}
Keep destructive tools such as cancel, pause, queue drop, daemon config-set,
plugin install, and plugin uninstall on manual approval unless the project
policy explicitly allows them.
Run separate MCP servers with distinct names:
{
"mcpServers": {
"animus-frontend": {
"command": "animus",
"args": ["--project-root", "/path/to/frontend", "mcp", "serve"]
},
"animus-backend": {
"command": "animus",
"args": ["--project-root", "/path/to/backend", "mcp", "serve"]
}
}
}
Tool ids include the server name, for example
mcp__animus-frontend__animus_subject_list.
animus --project-root /path/to/project mcp serve
The server speaks JSON-RPC 2.0 over stdio.
animus mcp memory starts the separate memory-context MCP server that
workflow phases use.
For MCP servers that require OAuth (configured with an oauth: block in
workflow config):
animus mcp auth <server> # browser login; tokens go to the OS keychain
animus mcp auth-status # show authenticated servers and token expiry
animus mcp auth-logout <server> # delete stored tokens
At run time Animus rewrites these servers to the animus-mcp-proxy stdio
bridge automatically, so tokens never reach CLI configs, .mcp.json, or
argv. Details live in docs/reference/mcp-oauth.md in the animus-cli repo.
animus binary path in .mcp.json is correct.animus --project-root . mcp serve starts without errors..mcp.json.enableAllProjectMcpServers: true in Claude settings.animus; many clients cache the tool list.animus.interactions.* only appear with --management.remediation.install_command — it names the exact fix.animus plugin install-defaults.animus daemon preflight and install any reported missing plugins.--project-root points to a directory containing .animus/ or a git repo.animus.tools.search with intent keywords, or animus.tools.list
for the full live catalog — results always reflect the serving binary.docs/reference/mcp-tools.md in the animus-cli repo.animus.task.* / animus.requirements.* calls to
animus.subject.*, and drop stale animus.runner.* calls (removed).