MCP Server Reference
The Crux MCP server is a single 6.6MB Rust binary (crux mcp start) exposing 66 tools via the Model Context Protocol. Any MCP-compatible AI assistant connects with one config line.
Connection
{
"type": "stdio",
"command": "/path/to/crux",
"args": ["mcp", "start"]
}
Session Tools
| Tool |
Description |
get_session_state |
Current mode, tool, working_on, decisions |
update_session |
Update working_on, add decision, add file |
restore_context |
Restore full session context — call at session start |
write_handoff |
Write handoff for next session |
read_handoff |
Read handoff from previous session |
recover_session |
Recover context from corrupted Claude Code sessions |
Knowledge & Memory
| Tool |
Description |
lookup_knowledge |
Search knowledge entries by keyword |
promote_knowledge |
Promote project knowledge to user scope |
remember_fact |
Remember a fact for future sessions |
recall_memories |
Search stored memories |
forget_fact |
Remove a stored memory |
list_all_memories |
List all stored memories |
Impact Analysis
| Tool |
Description |
analyze_impact |
Rank files by relevance (git + grep + AST) |
search_code |
Search codebase for files and symbols |
index_codebase |
Build/refresh the codebase index |
Git Context
| Tool |
Description |
git_context |
File history, risk score |
git_diff |
Current uncommitted changes |
git_risky_files |
Files with highest churn |
git_suggest_commit |
Suggest commit message from staged changes |
Safety Pipeline
| Tool |
Description |
validate_script |
Gate 1: preflight validation |
get_active_gates |
Active gates for a risk level |
get_pipeline_config |
Current pipeline configuration |
start_tdd_gate |
Gate 2: start TDD enforcement |
check_tdd_status |
TDD gate status |
start_security_audit |
Gate 3: start security audit |
security_audit_summary |
Security audit results |
audit_script_8b |
Gate 4: 8B adversarial audit (Ollama) |
audit_script_32b |
Gate 5: 32B second opinion (Ollama) |
Design Validation
| Tool |
Description |
start_design_validation |
WCAG/brand checks |
design_validation_summary |
Validation results |
check_contrast |
WCAG contrast ratio for two colors |
Modes
| Tool |
Description |
get_mode_prompt |
Full prompt text for a mode |
list_modes |
All available modes |
Logging
| Tool |
Description |
log_correction |
Log a correction for learning |
log_interaction |
Log a conversation message |
Tool Switching
| Tool |
Description |
switch_tool_to |
Switch to another AI tool |
MCP Server Registry
| Tool |
Description |
register_mcp_server |
Register an external MCP server |
remove_mcp_server |
Remove a server |
list_mcp_servers |
List registered servers |
Build-in-Public
| Tool |
Description |
bip_generate |
Check triggers, gather content for a draft |
bip_approve |
Approve and queue to Typefully |
bip_status |
Current BIP state and counters |
bip_get_analytics |
Engagement analytics |
Model Routing
| Tool |
Description |
get_model_for_task |
Recommended model for a task type |
get_available_tiers |
Available models at each tier (auto-discovers Ollama) |
get_mode_model |
Recommended model for a mode |
get_model_quality_stats |
Success rates per model per task |
Figma Integration
| Tool |
Description |
figma_get_tokens |
Extract design tokens from Figma |
figma_get_components |
Get component library |
Cross-Project
| Tool |
Description |
register_project |
Register for cross-project analytics |
get_cross_project_digest |
Digest spanning all projects |
Background Processors
| Tool |
Description |
check_processor_thresholds |
Which thresholds are exceeded |
run_background_processors |
Run due processors |
get_processor_status |
When each processor last ran |
Diagnostics
| Tool |
Description |
verify_health |
Health checks report |
get_project_context |
Auto-generated PROJECT.md |
get_digest |
Daily digest |