Menü
Reference for all Mira MCP tool signatures, parameters, and workflows.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
This skill should be used when the user asks "mira status", "show status", "how is mira doing", "system status", "index stats", "show mira stats", or wants a quick overview of Mira's index, storage, and active goals.
This skill should be used when the user asks "show my goals", "what should I work on", "update goal progress", "create a goal", "add milestone", "complete milestone", "track progress", "manage milestones", or wants to track work across multiple sessions.
This skill should be used when the user asks "what did you learn", "show insights", "any patterns noticed", "what should I know", "background analysis", "what patterns", or wants to see Mira's background analysis and predictions.
This skill should be used when the user asks "what were we working on", "session recap", "remind me of context", "catch me up", "restore context", "previous session", or starts a new session needing prior context.
This skill should be used when the user asks "analyze my changes", "what did I change", "review my diff", "impact of changes", "show changes", "git analysis", or wants semantic analysis of code changes.
This skill should be used when the user asks "find code that does X", "search for X", "where is X implemented", "what handles X", "find the X logic", "search by meaning", "semantic search", or needs to find code by concept or intent rather than exact text matching.
| name | tools-reference |
| description | Reference for all Mira MCP tool signatures, parameters, and workflows. |
Mira exposes 8 MCP tools to Claude Code. Additional actions are available via the mira tool CLI.
project — Project/Session Managementproject(action="start", project_path="...", name="...") # Initialize session
project(action="get") # Show current project
code — Code Intelligencecode(action="search", query="authentication middleware", limit=10)
code(action="symbols", file_path="src/main.rs", symbol_type="function")
code(action="callers", function_name="handle_login", limit=20)
code(action="callees", function_name="process_request", limit=20)
diff — Semantic Diff Analysisdiff() # Auto-detects staged/working/last commit
diff(from_ref="main", to_ref="feature/x")
diff(from_ref="v1.0", to_ref="v1.1", include_impact=true)
session — Session Managementsession(action="current_session") # Show current session
session(action="recap") # Quick overview: goals, sessions, insights
insights — Background Analysis & Insightsinsights(action="insights", insight_source="pondering", min_confidence=0.5)
insights(action="dismiss_insight", insight_id=42) # Remove resolved insight
goal — Cross-Session Goal Trackinggoal(action="create", title="...", description="...", priority="high")
goal(action="bulk_create", goals='[{"title": "...", "priority": "medium"}]')
goal(action="list", include_finished=false, limit=10)
goal(action="get", goal_id=1)
goal(action="update", goal_id=1, status="in_progress")
goal(action="delete", goal_id=1)
goal(action="add_milestone", goal_id=1, milestone_title="...", weight=2)
goal(action="complete_milestone", milestone_id=1)
goal(action="delete_milestone", milestone_id=1)
goal(action="progress", goal_id=1)
index — Code Indexingindex(action="project") # Full project index (auto-enqueues as background task)
index(action="project", skip_embed=true) # Fast re-index without embeddings
index(action="file", path="src/main.rs")
index(action="status") # Show index statistics
launch — Context-Aware Team Launcherlaunch(team="expert-review-team") # Parse agent file, enrich with project context
launch(team="expert-review-team", members="nadia,jiro") # Filter to specific members
launch(team="qa-hardening-team", scope="src/tools/") # Scope code context to a path
launch(team="refactor-team", context_budget=8000) # Custom context budget (default: 4000)
Returns LaunchData with pre-assembled agent specs (name, role, model, prompt, task_subject, task_description), shared project context, and suggested team ID. Does not spawn agents -- Claude orchestrates with TeamCreate/TaskCreate/Task.
The following actions are available via mira tool <name> '<json>' but not exposed as MCP tools:
| Tool | Action | Purpose |
|---|---|---|
project | set | Change active project without full init |
memory | export_claude_local | Export memories to CLAUDE.local.md |
code | dependencies | Module dependency graph |
code | patterns | Architectural pattern detection |
code | tech_debt | Per-module tech debt scores |
index | compact | VACUUM vec tables |
index | summarize | Generate module summaries |
index | health | Full code health scan |
session | list_sessions | List recent sessions |
session | get_history | View session tool history |
session | usage_summary | LLM usage totals |
session | usage_stats | LLM usage by dimension |
session | usage_list | Recent LLM usage records |
session | tasks_list | List background tasks |
session | tasks_get | Get background task status |
session | tasks_cancel | Cancel a background task |
session | storage_status | Database size and retention |
session | cleanup | Run data cleanup |
documentation | (all actions) | Documentation gap management |
team | (all actions) | Agent Teams intelligence |