MCP Server

The Dexicon MCP (Model Context Protocol) server lets AI assistants search your Dexicon knowledge base directly.

Installation

Connect Dexicon's MCP server to give your AI coding assistant access to your knowledge base.

MCP URL: https://api.dexicon.ai/mcp/

Cursor

Add to your Cursor MCP settings (.cursor/mcp.json):

json

{
  "mcpServers": {
    "dexicon": {
      "url": "https://api.dexicon.ai/mcp/",
      "transport": "sse",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Claude Code

Run this command in your terminal:

bash

claude mcp add --transport http dexicon https://api.dexicon.ai/mcp/ \
  --header "Authorization: Bearer YOUR_API_KEY"

If the command above doesn't work, you can manually add the following to your Claude settings file (~/.claude.json):

json

{
  "mcpServers": {
    "dexicon": {
      "type": "http",
      "url": "https://api.dexicon.ai/mcp/",
      "transport": "http",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

VS Code

Add to your VS Code MCP config (.vscode/mcp.json):

json

{
  "servers": {
    "dexicon": {
      "type": "sse",
      "url": "https://api.dexicon.ai/mcp/",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Codex

Add to your Codex config (~/.codex/config.toml):

ini

[mcp_servers.dexicon]
url = "https://api.dexicon.ai/mcp/"
http_headers = { "Authorization" = "Bearer YOUR_API_KEY" }

GitHub Copilot

Add to your GitHub Copilot MCP config (~/.copilot/mcp-config.json):

json

{
  "mcpServers": {
    "dexicon": {
      "type": "http",
      "url": "https://api.dexicon.ai/mcp/http",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      },
      "tools": ["*"]
    }
  }
}

MCP Tools Reference

Search & Discovery

search - Semantic search across knowledge, sessions, and documents

Searches knowledge first and automatically falls back to sessions and documents when knowledge results are insufficient. Returns results organized by type.

Parameters:

Parameter	Type	Required	Default	Description
`query`	string	Yes*	-	Natural language search query. Required for new searches.
`limit`	number	No	`5`	Results per page (1-20).
`include_summary`	boolean	No	`true`	Include full summary text (100-250 words).
`tags`	string[]	No	-	Filter by tags (OR logic). Get valid tags from `list_work_topics`.
`lookback_days`	number	No	-	Number of days to look back from now. Mutually exclusive with `start_time`/`end_time`.
`start_time`	string	No	-	Filter results after this time (ISO 8601 UTC).
`end_time`	string	No	-	Filter results before this time (ISO 8601 UTC).
`project_id`	string	No	-	Scope search to a specific codebase. Get IDs from `list_projects`.
`email`	string	No	-	Filter to a specific user's sessions by email address (case-insensitive).
`knowledge_type`	string	No	-	Filter knowledge by type: `general`, `bug`, `standard`, or `decision`.
`cursor`	string	No	-	Opaque cursor from previous response for pagination.
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "knowledge": [
    {
      "id": "mem_01ABC...",
      "title": "PostgreSQL connection pooling best practices",
      "content": "Always use connection pooling...",
      "context": "When setting up database connections...",
      "tags": ["database", "postgresql"],
      "importance": 8,
      "similarity": 0.92,
      "created_at": "2025-01-15T10:00:00+00:00"
    }
  ],
  "sessions": [
    {
      "source": "sessions",
      "source_id": "ccs_01ABC...",
      "summary_id": "ssm_01XYZ...",
      "title": "Implemented OAuth2 authentication",
      "preview": "Added JWT token handling...",
      "summary": "Full summary text if include_summary=true...",
      "technologies": ["Python", "FastAPI"],
      "tags": ["auth", "security"],
      "user": {
        "email": "developer@company.com",
        "name": null
      },
      "similarity": 0.87,
      "created_at": "2025-01-15T10:00:00+00:00"
    }
  ],
  "documents": [
    {
      "source": "documents",
      "source_id": "doc_01ABC...",
      "title": "Architecture Overview",
      "preview": "System design document...",
      "similarity": 0.85,
      "created_at": "2025-01-15T10:00:00+00:00"
    }
  ],
  "returned_count": 5,
  "next_cursor": "eyJxdWVyeSI6...",
  "query": "authentication",
  "fallback_triggered": false
}

list_work_topics - Discover what tags exist in recent coding sessions

Call this FIRST before using get_work_by_topic to see valid tags. Returns tags sorted by frequency (most common first).

Parameters:

Parameter	Type	Required	Default	Description
`cursor`	string	No	-	Opaque cursor from previous response for pagination.
`lookback_days`	number	No	`30`	How far back to look (1-90 days).
`limit`	number	No	`20`	Results per page (1-50).
`scope`	string	No	`"team"`	`"me"` (your sessions only) or `"team"` (everyone in org).
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "topics": [
    {
      "tag": "authentication",
      "session_count": 12
    },
    {
      "tag": "database",
      "session_count": 8
    }
  ],
  "returned_count": 10,
  "next_cursor": "eyJvIjoy...",
  "lookback_days": 30,
  "scope": "team"
}

get_work_by_topic - Get sessions filtered by a specific topic tag

PRIMARY tool for browsing work by category. Faster than search when you know the topic. Workflow: list_work_topics → get_work_by_topic → get_session.

Parameters:

Parameter	Type	Required	Default	Description
`tag`	string	Yes	-	Topic to filter by. Get valid tags from `list_work_topics` first.
`cursor`	string	No	-	Opaque cursor from previous response for pagination.
`lookback_days`	number	No	`30`	How far back to look (1-90 days).
`limit`	number	No	`10`	Results per page (1-50).
`scope`	string	No	`"team"`	`"me"` (your sessions only) or `"team"` (everyone in org).
`email`	string	No	-	Filter to a specific user's sessions by email address (case-insensitive). Overrides `scope`.
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "sessions": [
    {
      "id": "ssm_01ABC...",
      "session_id": "ccs_01XYZ...",
      "provider": "claude_code",
      "title": "Implemented OAuth2 authentication",
      "preview": "Added JWT token handling and refresh flow...",
      "tags": ["authentication", "security"],
      "technologies": ["Python", "FastAPI"],
      "outcome": "completed",
      "user": {
        "email": "developer@company.com",
        "name": null
      },
      "created_at": "2025-01-15T10:00:00+00:00"
    }
  ],
  "returned_count": 5,
  "next_cursor": "eyJvIjoy...",
  "tag": "authentication",
  "lookback_days": 30,
  "scope": "team"
}

get_recent_work - Get recent sessions sorted by time (newest first)

Chronological view across all topics. For topic-specific browsing, use get_work_by_topic instead.

Parameters:

Parameter	Type	Required	Default	Description
`cursor`	string	No	-	Opaque cursor from previous response for pagination.
`lookback_days`	number	No	`30`	How far back to look (1-90 days).
`limit`	number	No	`10`	Results per page (1-50).
`scope`	string	No	`"team"`	`"me"` (your sessions only) or `"team"` (everyone in org).
`email`	string	No	-	Filter to a specific user's sessions by email address (case-insensitive). Overrides `scope`.
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "sessions": [
    {
      "id": "ssm_01ABC...",
      "session_id": "ccs_01XYZ...",
      "provider": "claude_code",
      "title": "Implemented OAuth2 authentication",
      "preview": "Added JWT token handling and refresh flow...",
      "tags": ["authentication", "security"],
      "technologies": ["Python", "FastAPI"],
      "outcome": "completed",
      "user": {
        "email": "developer@company.com",
        "name": null
      },
      "created_at": "2025-01-15T10:00:00+00:00"
    }
  ],
  "returned_count": 10,
  "next_cursor": "eyJvIjoy...",
  "lookback_days": 30,
  "scope": "team"
}

Sessions & Documents

get_session - Fetch full conversation transcript with all messages

Use after finding sessions via search, get_work_by_topic, or get_recent_work to retrieve the actual messages (human + assistant), not just the summary.

Parameters:

Parameter	Type	Required	Default	Description
`session_id`	string	Yes	-	Session ID from search/browse results. Prefix indicates provider: `ccs_` (Claude Code), `curss_` (Cursor), `ghcs_*` (GitHub Copilot).
`cursor`	string	No	-	Pagination cursor from previous response.
`limit`	number	No	`100`	Messages per page (1-200).
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "id": "ccs_01ABC...",
  "provider": "claude_code",
  "title": "Implemented OAuth2 authentication",
  "summary": "Full session summary text...",
  "project": {
    "id": "ccp_01XYZ...",
    "name": "my-project"
  },
  "user": {
    "email": "developer@company.com",
    "name": "Kevin"
  },
  "total_messages": 47,
  "messages": [
    {
      "id": "ccm_01ABC...",
      "role": "user",
      "content": "Help me implement OAuth2...",
      "timestamp": "2025-01-15T10:00:00+00:00"
    },
    {
      "id": "ccm_01DEF...",
      "role": "assistant",
      "content": "I'll help you implement OAuth2...",
      "timestamp": "2025-01-15T10:00:05+00:00"
    }
  ],
  "next_cursor": "eyJpZCI6...",
  "has_more": true
}

get_document - Fetch full document content by ID

Use this after search to retrieve the complete text of a document.

Parameters:

Parameter	Type	Required	Description
`document_id`	string	Yes	The document UUID from search results.
`api_key`	string	No	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "document_id": "doc_01ABC...",
  "original_name": "architecture.md",
  "full_text": "Complete document content...",
  "mime_type": "text/markdown",
  "tags": ["architecture", "design"],
  "created_at": "2025-01-15T10:00:00+00:00"
}

list_projects - List projects (codebases) to scope searches

Get project IDs for filtering search to specific repositories.

Parameters:

Parameter	Type	Required	Default	Description
`cursor`	string	No	-	Opaque cursor from previous response for pagination.
`only_my_recent_projects`	boolean	No	`false`	Only projects you've worked on recently. Good for "what was I working on?" queries.
`lookback_days`	number	No	`30`	How far back (1-90) when `only_my_recent_projects=true`.
`limit`	number	No	`20`	Results per page (1-100).
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "projects": [
    {
      "project_id": "ccp_01ABC...",
      "name": "my-project",
      "updated_at": "2025-01-15T10:00:00+00:00",
      "session_count": 15,
      "last_ide_activity_at": "2025-01-15T09:30:00+00:00"
    }
  ],
  "returned_count": 10,
  "next_cursor": "eyJvIjoy...",
  "lookback_days": 30,
  "only_my_recent_projects": false
}

Knowledge Management

create_knowledge - Create a new knowledge item

Create a new knowledge item in the knowledge base. Knowledge items store reusable information for AI agents (patterns, decisions, debugging tips, etc.). They support semantic search and knowledge graph linking.

Parameters:

Parameter	Type	Required	Default	Description
`title`	string	Yes	-	Descriptive title (max 250 chars).
`content`	string	Yes	-	The actual knowledge content (~400 words max recommended).
`context`	string	Yes	-	WHY/HOW/WHAT context explaining when to use this knowledge.
`keywords`	string[]	No	-	Search terms in slug format (lowercase, alphanumeric, dashes).
`tags`	string[]	No	-	Category tags in slug format (max 10).
`importance`	number	No	`7`	Priority 1-10. Higher = more prominent in search.
`visibility`	string	No	`"organization"`	`"organization"` (everyone) or `"user"` (private).
`linked_memory_ids`	string[]	No	-	IDs of related knowledge items for knowledge graph linking.
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "id": "mem_01ABC...",
  "organization_id": "org_01XYZ...",
  "user_id": "usr_01DEF...",
  "title": "PostgreSQL connection pooling best practices",
  "content": "Always use connection pooling with PgBouncer...",
  "context": "When setting up database connections in production...",
  "keywords": ["postgresql", "connection-pooling", "pgbouncer"],
  "tags": ["database", "performance"],
  "importance": 8,
  "visibility": "organization",
  "creator_type": "agent",
  "linked_memory_ids": [],
  "source_type": null,
  "source_id": null,
  "memory_type": "general",
  "is_obsolete": false,
  "created_at": "2025-01-15T10:00:00+00:00",
  "updated_at": "2025-01-15T10:00:00+00:00"
}

get_knowledge - Get a knowledge item by ID

Retrieve full details of a specific knowledge item including content, metadata, and linked items.

Parameters:

Parameter	Type	Required	Description
`memory_id`	string	Yes	The knowledge item ID (`mem_*` prefix).
`api_key`	string	No	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "id": "mem_01ABC...",
  "organization_id": "org_01XYZ...",
  "user_id": "usr_01DEF...",
  "title": "PostgreSQL connection pooling best practices",
  "content": "Always use connection pooling with PgBouncer...",
  "context": "When setting up database connections in production...",
  "keywords": ["postgresql", "connection-pooling", "pgbouncer"],
  "tags": ["database", "performance"],
  "importance": 8,
  "visibility": "organization",
  "creator_type": "agent",
  "linked_memory_ids": ["mem_02XYZ..."],
  "source_type": null,
  "source_id": null,
  "memory_type": "general",
  "is_obsolete": false,
  "created_at": "2025-01-15T10:00:00+00:00",
  "updated_at": "2025-01-15T10:00:00+00:00"
}

list_knowledge - List knowledge items with filters

Browse knowledge by tags, keywords, importance, or creator type. Results are ordered by importance (high first), then by creation time.

Parameters:

Parameter	Type	Required	Default	Description
`cursor`	string	No	-	Opaque cursor from previous response for pagination.
`limit`	number	No	`10`	Results per page (1-50).
`tags`	string[]	No	-	Filter by tags (OR logic).
`keywords`	string[]	No	-	Filter by keywords (OR logic).
`min_importance`	number	No	-	Only return items with importance >= this value.
`include_obsolete`	boolean	No	`false`	Include obsolete items.
`creator_type`	string	No	-	Filter by `"agent"` or `"user"`.
`source_type`	string	No	-	Filter by source type: `"document"` or `"session"`.
`source_id`	string	No	-	Filter by source ID (document UUID or session ID).
`memory_type`	string	No	-	Filter by type: `"general"`, `"bug"`, `"standard"`, `"decision"`.
`api_key`	string	No	-	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "items": [
    {
      "id": "mem_01ABC...",
      "title": "PostgreSQL connection pooling best practices",
      "content": "Always use connection pooling...",
      "context": "When setting up database connections...",
      "tags": ["database", "performance"],
      "importance": 8,
      "memory_type": "general",
      "created_at": "2025-01-15T10:00:00+00:00"
    }
  ],
  "total": 25,
  "returned_count": 10,
  "next_cursor": "eyJvIjoy..."
}

update_knowledge - Update a knowledge item

Update a knowledge item (PATCH semantics - only provided fields are updated). Only the owner can update. If content, context, or title change, the embedding is regenerated for search.

Parameters:

Parameter	Type	Required	Description
`memory_id`	string	Yes	The knowledge item ID to update (`mem_*` prefix).
`title`	string	No	New title (max 250 chars).
`content`	string	No	New content.
`context`	string	No	New context.
`keywords`	string[]	No	New keywords list.
`tags`	string[]	No	New tags list.
`importance`	number	No	New importance (1-10).
`api_key`	string	No	API key for authentication. Use when Authorization header is not available.

Response:

Returns the updated knowledge item with the same structure as create_knowledge.

delete_knowledge - Delete a knowledge item

Soft delete a knowledge item. Only the owner can delete. The item can be restored by admins.

Parameters:

Parameter	Type	Required	Description
`memory_id`	string	Yes	The knowledge item ID to delete (`mem_*` prefix).
`api_key`	string	No	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "status": "deleted",
  "memory_id": "mem_01ABC..."
}

list_knowledge_tags - Get tag counts from knowledge items

Returns a dictionary mapping tags to their occurrence count. Useful for discovering what knowledge categories exist.

Parameters:

Parameter	Type	Required	Description
`api_key`	string	No	API key for authentication. Use when Authorization header is not available.

Response:

json

{
  "tags": {
    "database": 15,
    "authentication": 12,
    "performance": 8,
    "api": 6
  }
}

MCP Server ​

Installation ​

MCP Tools Reference ​

Search & Discovery ​

Sessions & Documents ​

Knowledge Management ​

MCP Server

Installation

MCP Tools Reference

Search & Discovery

Sessions & Documents

Knowledge Management