Msty Admin MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MSTY_HOST | No | Msty installation host | 127.0.0.1 |
| MSTY_AI_PORT | No | Local AI (Ollama) port | 11964 |
| MSTY_TIMEOUT | No | Service timeout in seconds | 10 |
| MSTY_MLX_PORT | No | MLX service port | 11973 |
| MSTY_VIBE_PORT | No | Vibe CLI Proxy port | 8317 |
| ANTHROPIC_API_KEY | No | Required for Bloom judge model (Phase 6) | |
| MSTY_LLAMACPP_PORT | No | LLaMA.cpp service port | 11454 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| detect_msty_installationB | Detect Msty Studio installation, version, database, and service ports. |
| read_msty_databaseB | Query the Studio workspace database. Use query='stats' for overview, query='tables' for table list, or a read-only SQL SELECT/WITH/PRAGMA statement. |
| list_configured_toolsA | List MCP/Toolbox tools configured in Msty Studio. |
| get_model_providersA | List local service backends and Studio-configured providers (keys redacted). |
| analyse_msty_healthB | Comprehensive health report for Studio install, DB, and services. |
| get_server_statusB | MCP server status and tool inventory summary. |
| export_tool_configC | Export a Studio tool configuration by name or toolId. |
| sync_claude_preferencesB | Report Claude Desktop MCP sync candidates (read-only inventory). |
| generate_personaC | Generate a persona definition JSON suitable for Persona Studio import. |
| import_tool_configA | Validate a tool config payload for Toolbox import. Live SQLite writes are refused unless confirm=true (still not recommended while Studio is open). |
| get_service_statusB | Status of Local AI, MLX, Llama.cpp, Vibe, and Nexus ports. |
| list_available_modelsB | List models across local services. |
| query_local_ai_serviceD | Query Local AI (Ollama-compatible) service. |
| chat_with_local_modelC | Chat with a Local AI model. |
| recommend_modelC | Recommend a model using Studio insights when available. |
| list_mlx_modelsC | List MLX models from the MLX service. |
| chat_with_mlx_modelC | Chat with an MLX model. |
| list_llamacpp_modelsC | List Llama.cpp models. |
| chat_with_llamacpp_modelC | Chat with a Llama.cpp model. |
| get_vibe_proxy_statusB | Check Vibe CLI proxy status. |
| query_vibe_proxyD | Query Vibe CLI proxy. |
| get_model_performance_metricsC | Performance metrics from Studio insightsLanguageModelUsage. |
| analyse_conversation_patternsC | Analyse conversation patterns from Studio data. |
| compare_model_responsesC | Compare responses from multiple local models (best-effort). |
| optimise_knowledge_stacksB | Suggest Knowledge Stack optimisations from live Studio data. |
| suggest_persona_improvementsC | Suggest persona improvements from live Studio personas. |
| run_calibration_testC | Run calibration test on a local model. |
| evaluate_response_qualityC | Evaluate response quality with heuristic rubric. |
| identify_handoff_triggersB | Identify handoff patterns from insights + calibration heuristics. |
| get_calibration_historyC | Get calibration history from ~/.msty-admin metrics DB. |
| bloom_evaluate_modelC | Run Bloom behavioural evaluation on a model. |
| bloom_check_handoffC | Check whether a model should hand off to Claude for a task category. |
| bloom_get_historyC | Get Bloom evaluation history. |
| bloom_list_behaviorsB | List available Bloom behaviours. |
| bloom_get_thresholdsC | Get quality thresholds for a task category. |
| bloom_validate_modelC | Validate model suitability for Bloom evaluation. |
| list_studio_catalogueB | Summary counts for all known Studio entity types. |
| list_workspacesC | List Studio workspaces. |
| list_personasC | List personas from Persona Studio. |
| list_shadow_personasD | List Shadow Personas. |
| list_crewsC | List Crew Mode crews. |
| list_knowledge_stacksC | List Knowledge Stacks (including Next Gen). |
| list_turnstilesC | List Turnstiles / Turnstiles vNext definitions. |
| list_skillsC | List Skills Studio preferences and folders. |
| list_live_contextsC | List Live Contexts. |
| list_promptsC | List Prompts Studio library entries. |
| list_context_studioC | List Context Studio attachments and folders (Studio 2.9). |
| list_agent_plansD | List Agent Mode plans. |
| get_studio_entityA | Fetch a single Studio entity by type and id.
|
| get_schema_migrationsB | Show recent Drizzle schema migrations from the Studio database. |
| export_workflow_packC | Export a read-only workflow pack (personas, tools, turnstiles, stacks, …). |
| detect_nexusC | Detect Msty Nexus local gateway. |
| list_nexus_modelsB | List models exposed by the Nexus OpenAI-compatible endpoint. |
| query_nexusC | Query the Nexus gateway with an arbitrary path. |
| get_insights_usageD | Alias for insights-backed usage analytics. |
| ask_frontierC | Answer a hard question. Call once, then stop. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 56 tools
Many tools target distinct resources (Bloom evaluations, Studio entities, model services, Nexus), but several clusters overlap: get_service_status vs analyse_msty_health vs get_server_status, multiple chat_with_*_model variants, and multiple list_*_models tools. Descriptions help, but boundaries remain fuzzy for status/health and model-selection tasks.
Names are predominantly snake_case with verb_noun or verb_entity order, and spelling is consistently British (analyse, optimise). Minor deviations exist: bloom_* prefixes some Bloom tools, and retrieval verbs vary across get, read, and query.
With 56 tools, the set is far beyond the typical 3-15 well-scoped range. The breadth includes many niche, overlapping, and read-only inspection tools, making the surface difficult for an agent to navigate efficiently.
The server covers a wide read-only surface: installation health, databases, services, models, Bloom evaluations, Studio entities, migrations, and Nexus. However, mutation operations (create/update/delete) are largely absent for Studio entities, and import/sync tools only validate or report rather than perform changes, creating admin dead ends.