Skip to main content
Glama
spirrow-prismind

spirrow-prismind

Official
by SpirrowGames

Spirrow-Prismind

License: MIT Python 3.11+ MCP

MCP Server for Context-Aware Knowledge Management

An MCP (Model Context Protocol) server that integrates multiple data sources—Google Drive, RAG (Retrieval-Augmented Generation), and MCP Memory Server—to provide context-aware knowledge management for AI assistants.

Note: 日本語版は README.ja.md を参照してください。

Features

  • Session Management: Save and restore work states, automatically suggest relevant documents

  • Project Management: Switch between multiple projects, search for similar projects

  • Document Operations: Google Docs integration with automatic catalog registration

  • Catalog Management: Semantic search with Google Sheets synchronization

  • Knowledge Management: RAG-based knowledge storage and retrieval

Related MCP server: brkt-gdrive-mcp

Requirements

  • Python 3.11 or higher

  • Google Cloud project with OAuth 2.0 credentials

  • RAG server (ChromaDB compatible)

  • MCP Memory Server (optional)

Installation

# Clone the repository
git clone https://github.com/SpirrowGames/spirrow-prismind.git
cd spirrow-prismind

# Install the package
pip install -e .

Configuration

  1. Copy the example config:

    cp config.toml.example config.toml
  2. Set up Google OAuth credentials:

    • Create a project in Google Cloud Console

    • Enable Google Docs, Drive, and Sheets APIs

    • Create OAuth 2.0 credentials (Desktop application)

    • Download credentials.json and place it in the config directory

  3. Configure external services in config.toml:

    • RAG Server endpoint (ChromaDB compatible)

    • MCP Memory Server connection (optional)

Usage

With Claude Desktop

Add to your Claude Desktop configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "prismind": {
      "command": "spirrow-prismind",
      "env": {
        "PRISMIND_CONFIG": "/path/to/config.toml"
      }
    }
  }
}

With Claude Code

Add to your Claude Code settings:

{
  "mcpServers": {
    "prismind": {
      "command": "spirrow-prismind",
      "env": {
        "PRISMIND_CONFIG": "/path/to/config.toml"
      }
    }
  }
}

Available Tools

Session Management

Tool

Description

start_session

Start a session and restore previous state

end_session

End session and save state for handoff

save_session

Save intermediate session state

list_sessions

List all sessions for a project

delete_session

Delete a specific session

list_context_authors

List the distinct context authors/roles saved for a project

start_session / end_session / save_session / delete_session accept an optional author argument. It adds a partition segment to the session key (prismind:session:{project}:{user}:{author}), so one project+user can hold multiple independent contexts — one per author/role. An empty author keeps the legacy key for backward compatibility.

Project Management

Tool

Description

setup_project

Set up a new project

switch_project

Switch to another project

list_projects

List all projects

update_project

Update project settings

delete_project

Delete a project

Document Operations

Tool

Description

get_document

Search and retrieve documents

create_document

Create a new document

update_document

Update an existing document

Catalog Operations

Tool

Description

search_catalog

Search the document catalog

sync_catalog

Sync catalog (Sheets → RAG)

Knowledge Operations

Tool

Description

add_knowledge

Add knowledge entries

search_knowledge

Search knowledge base

update_knowledge

Update an existing knowledge entry

delete_knowledge

Delete a knowledge entry

Progress Management

Tool

Description

get_progress

Get project progress with phases and tasks

add_task

Add a new task to progress sheet

get_task

Get single task details by ID

update_task

Update task (name, description, status, priority, phase move)

delete_task

Delete task with blocked_by reference cleanup

update_task_status

Update task status

start_task

Mark task as in_progress

complete_task

Mark task as completed

block_task

Mark task as blocked

Architecture

┌──────────────────────────────────────────────────────────────┐
│                       MCP Server                             │
│  ┌────────────────────────────────────────────────────────┐  │
│  │                        Tools                           │  │
│  │  ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐      │  │
│  │  │ Session │ │ Project │ │Document │ │Knowledge│ ...  │  │
│  │  └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘      │  │
│  └───────┼──────────┼──────────┼──────────┼─────────────┘  │
│          │          │          │          │                 │
│  ┌───────┴──────────┴──────────┴──────────┴─────────────┐  │
│  │                    Integrations                       │  │
│  │  ┌────────┐ ┌────────┐ ┌────────┐ ┌────────┐         │  │
│  │  │  RAG   │ │ Memory │ │  Docs  │ │ Drive  │         │  │
│  │  │ Client │ │ Client │ │ Client │ │ Client │         │  │
│  │  └────┬───┘ └────┬───┘ └────┬───┘ └────┬───┘         │  │
│  └───────┼──────────┼──────────┼──────────┼─────────────┘  │
└──────────┼──────────┼──────────┼──────────┼────────────────┘
           │          │          │          │
           ▼          ▼          ▼          ▼
      ┌────────┐ ┌────────┐ ┌────────────────────┐
      │ChromaDB│ │ Memory │ │    Google APIs     │
      │  RAG   │ │ Server │ │ Docs/Drive/Sheets  │
      └────────┘ └────────┘ └────────────────────┘

Development

Setup

# Install with dev dependencies
pip install -e ".[dev]"

Testing

pytest

Code Quality

# Format code
ruff format .

# Lint and fix
ruff check --fix .

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

  1. Fork the repository

  2. Create your feature branch (git checkout -b feature/amazing-feature)

  3. Commit your changes (git commit -m 'Add some amazing feature')

  4. Push to the branch (git push origin feature/amazing-feature)

  5. Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Available Tools

46 tools
add_knowledgeB

Add a knowledge entry to RAG. Store insights, tips, best practices, and lessons learned.

Inputs: content (required), category (required), tags, source, +1 more.

add_taskB

Add a new task to the progress sheet.

Inputs: name (required), phase (required), task_id (required), user, +5 more.

block_taskB

Mark a task as blocked with blockers list.

Inputs: phase (required), task_id (required), blockers (required), project.

check_services_statusB

Check connection status of RAG and Memory servers. Verifies server availability and collection/schema auto-creation status. Use detailed=true for protocol, latency, and other detailed info.

Inputs: detailed.

complete_taskB

Mark a task as completed. Convenience method that sets status to 'completed'.

Inputs: phase (required), task_id (required), notes, project.

configureA

Update Spirrow-Prismind configuration. Writes settings to config.toml.

Inputs: value (required), setting (required).

create_documentC

Create a new document and register it in the catalog.

Inputs: name (required), content (required), doc_type (required), phase_task (required), +3 more.

delete_documentA

Delete a document and its catalog entries. Requires project name to prevent accidental deletion.

Inputs: doc_id (required), project (required), soft_delete, delete_drive_file.

delete_document_typeA

Delete a document type. Use scope='global' for global types, or scope='project' for project-specific types.

Inputs: type_id (required), scope.

delete_knowledgeB

Delete a knowledge entry from RAG.

Inputs: knowledge_id (required), project.

delete_projectC

Delete project settings. Optionally delete the Google Drive folder as well.

Inputs: project (required), confirm, delete_drive_folder.

delete_sessionB

Delete a saved session state for a project.

Inputs: project (required), user, author.

delete_taskA

Delete a task from the progress sheet. Also cleans up blocked_by references in other tasks.

Inputs: phase (required), task_id (required), user, project.

end_sessionB

End the session and save state. Uses current project if project is not specified. Optionally declare 'embodiment' to record on the saved SessionState.

Inputs: notes, author, project, summary, +3 more.

export_server_configA

Export server configuration for team sharing. Outputs in TOML format excluding sensitive information (paths).

Inputs: no named parameters.

find_similar_document_typeA

Find a document type semantically similar to the query. Uses RAG-based semantic search (BGE-M3 embeddings) for multilingual matching. For example, 'api仕様' can match 'api_spec'.

Inputs: type_query (required), threshold.

get_connection_infoA

Get current connection information. Shows connection status, latency, and version info for Memory Server, RAG Server, and Google services.

Inputs: no named parameters.

get_documentC

Search and retrieve a document.

Inputs: query, doc_id, project, doc_type, +1 more.

get_identityA

Look up ONE identity record by name (cross-project, read-only). Use this -- not list_context_authors -- when you need a specific actor's allowed_roles / independence_class. list_context_authors is project-scoped and enumerates saved session state, so an identity that is registered but has never checkpointed in that project is absent from its output; treating that absence as 'unregistered' would silently skip Magickit's role x allowed_roles check. Response distinguishes 'not registered' (success=true, found=false) from 'lookup failed' (success=false) so callers can fail closed on the latter.

Inputs: identity_name (required), user.

get_progressC

Get project progress from Google Sheets.

Inputs: phase, project.

get_setup_statusA

Get the configuration status of Spirrow-Prismind. Shows required and optional settings with their configured/unconfigured state.

Inputs: no named parameters.

get_taskB

Get a single task by ID.

Inputs: phase (required), task_id (required), user, project.

import_server_configA

Import shared server configuration. Validates settings and reports any errors.

Inputs: config (required).

list_context_authorsA

List the distinct context authors/roles that have saved session state for a project. Use this to avoid creating duplicate contexts from naming variations and to check whether your own author's context already exists. Each entry carries an 'identity' object (or null) joined from the cross-project identity record (see upsert_identity), so callers see allowed_roles without a second lookup.

Inputs: project (required), user.

list_documentsB

List documents in a project with filtering and pagination.

Inputs: limit, offset, feature, project, +4 more.

list_document_typesA

Get a list of available document types. Returns both global types (shared across projects) and project-specific types. When same type_id exists in both, project type takes precedence.

Inputs: no named parameters.

list_projectsA

Get a list of registered projects.

Inputs: no named parameters.

list_sessionsA

List all saved sessions. Can filter by project or user.

Inputs: user, project.

register_document_typeA

Register a new document type. Use scope='global' for types shared across all projects, or scope='project' for project-specific types.

Inputs: name (required), type_id (required), folder_name (required), scope, +4 more.

save_sessionA

Save session state without ending the session. Uses current project if project is not specified. Optionally declare 'embodiment' to record on SessionState.

Inputs: notes, author, project, summary, +5 more.

search_catalogC

Search the document catalog.

Inputs: limit, query, feature, doc_type, +1 more.

search_knowledgeC

Search knowledge entries. Find relevant insights, tips, and lessons learned.

Inputs: query (required), tags, limit, project, +2 more.

setup_projectA

Set up a new project. If spreadsheet_id and root_folder_id are omitted, automatically creates them under the projects_folder_id configured in config.toml.

Inputs: name (required), project (required), force, description, +4 more.

start_sessionA

Start a session and load saved state. Uses current project if project is not specified. Specify 'author' to restore a specific context-author/role partition. Optionally declare 'embodiment' (ADR-2026-05-29-12 self-declared runtime form) to record on the loaded SessionState.

Inputs: author, project, embodiment.

start_taskA

Mark a task as in progress. Convenience method that sets status to 'in_progress'.

Inputs: phase (required), task_id (required), project.

switch_projectB

Switch to a different project.

Inputs: project (required).

sync_catalogB

Sync the catalog from Google Sheets to RAG cache.

Inputs: project.

sync_projects_from_driveA

Sync folder list under projects_folder_id in Google Drive with RAG. Uses Drive as master to add/remove projects.

Inputs: dry_run.

update_documentC

Update a document.

Inputs: doc_id (required), append, content, feature, +3 more.

update_knowledgeA

Update an existing knowledge entry.

Inputs: knowledge_id (required), tags, source, content, +1 more.

update_projectC

Update project settings.

Inputs: project (required), name, phases, status, +5 more.

update_session_progressC

Update session progress (phase, task, blockers). Uses current project if project is not specified. Optionally declare 'embodiment' to record on SessionState.

Inputs: author, project, blockers, embodiment, +3 more.

update_summaryB

Update the project summary sheet with information like description, current phase, and task counts.

Inputs: project, description, total_tasks, current_phase, +2 more.

update_taskA

Update a task with any combination of fields. Supports updating name, description, status, priority, category, blocked_by, blockers, and moving to a new phase.

Inputs: phase (required), task_id (required), name, user, +8 more.

update_task_statusC

Update task status in Google Sheets.

Inputs: phase (required), status (required), task_id (required), user, +6 more.

upsert_identityA

Create or update an identity record (cross-project actor declaration). Identity records persist allowed_roles / independence_class / persona_description in a key space separate from session state (prismind:identity:{user}:{identity_name}), realizing ADR-2026-05-27-09 D-3's persona-continuity gate at the API level. identity_name is the same string SessionState.author uses; list_context_authors joins the identity record onto each author entry so callers see allowed_roles in one round-trip. independence_class is required on every upsert (re-declared, not preserved) -- this is the '書き忘れ不能' guarantee. allowed_roles is required unless keep_allowed_roles=True. persona_description defaults to preserve-on-omit. embodiment is DEPRECATED (ADR-2026-05-29-12); runtime self-declared via the five APIs on Magickit (checkpoint/resume/chatroom_*). Pass omit or null; the field is preserved in the schema only for the staged migration window.

Inputs: identity_name (required), independence_class (required), user, embodiment, +3 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.3/5.0

    Scored across 46 tools

    Disambiguation3/5

    Most tools target distinct resources/actions, but there is overlap between start_task/complete_task/block_task and update_task_status, as well as between check_services_status and get_connection_info, which could cause misselection. Descriptions help clarify intent, but the boundaries are not always clean.

    Naming Consistency5/5

    All tools use snake_case with a clear verb_noun pattern (get_, list_, create_, update_, delete_, etc.). The verbs are consistent and predictable, with only minor special cases like 'configure' and 'upsert_identity'.

    Tool Count2/5

    With 46 tools, this is significantly more than typical and feels heavy. Many convenience wrappers (e.g., start_task, complete_task, block_task) could be folded into update_task_status, and several config/status tools could be consolidated.

    Completeness5/5

    The tool surface covers CRUD and lifecycle operations for sessions, projects, documents, knowledge, and tasks, plus configuration and identity management. No major gaps are apparent; the domain appears fully covered.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers