Factsets

Factsets MCP Server provides a persistent, self-maintaining knowledge base for AI agents that communicates via the Model Context Protocol. It organizes atomic facts, cached resources, procedural skills, and execution logs in SQLite and enables agents to access rich context across sessions with flexible tagging and preferences.

How to use

You can run Factsets as an MCP server and connect your MCP client to it to access persistent context, manage facts, resources, and skills, and track execution history. Use it to enrich conversations with structured knowledge and keep agent behavior consistent across sessions.

How to install

Prerequisites: ensure you have a modern Node.js runtime and a package manager available on your system (npm, pnpm, or bun). You will install Factsets globally, then run the MCP server.

Step-by-step setup and run

npm install --global factsets
pnpm install --global factsets
bun install --global factsets

# Start MCP server using the default command that auto-watches skills and seeds starter content
bunx factsets mcp-server

# Start MCP server with default command (alternate form)
bunx factsets

# Run without file watching
bunx factsets --no-watch-skills

# Run without seeding starter content
bunx factsets --no-seed

# Run file watcher standalone
bunx factsets watch-files --database-url <path>

# Run background maintenance worker
bunx factsets worker --database-url <path>

# Export database to JSON
bunx factsets dump backup.json

# Restore database from JSON
bunx factsets restore backup.json

# Example MCP client configuration snippet
{
  "mcpServers": {
    "factsets": {
      "command": "bunx",
      "args": ["factsets", "mcp-server"]
    }
  }
}

Connect to the MCP server from your client

Configure your MCP client to use the Factsets MCP server. Use the same command shown above to run the server, and reference the server in your client’s MCP configuration under mcpServers with the Bunx-based start command and arguments.

Core concepts you will use

Familiarize yourself with the core elements Factsets manages: Facts (atomic knowledge), Resources (cached external content), Skills (procedural markdown), Execution Logs (command history), and Tags (flexible categorization). Your agent will leverage these to build rich context and maintain state across interactions.

Tools and endpoints you will use

The server exposes tools to manage and query the MCP content. You can submit, search, verify, update, delete, and restore facts; manage resources and skills; record and query execution logs; and handle tags and configurations. The operator-level tools enable you to maintain a healthy knowledge base and coherent agent behavior.

Configuration and options

You can tailor how Factsets presents results and how the agent behaves through user preferences and a set of configuration options. Explore the available configuration prompts and guides to adjust output style, agent workflows, and context-building behavior.

Maintenance and diagnostics

Regularly check for stale resources and run maintenance routines to refresh content. Use the maintenance tools to generate reports, refresh guides, and ensure your facts and resources stay current.

Available tools

submit_facts

Add facts with tags and source tracking

search_facts

Query facts by tags, content, or filters

verify_facts

Mark facts as verified by ID

verify_facts_by_tags

Bulk verify facts by tags

update_fact

Update fact content, metadata, or tags

delete_facts

Remove facts by criteria

restore_facts

Restore soft-deleted facts

add_resources

Register resources with retrieval methods

search_resources

Find resources by tags, type, or URI

get_resources

Get resources by ID or URI with freshness

update_resource_snapshot

Update cached content for single resource

update_resource_snapshots

Bulk update cached content

update_resource

Update resource metadata (not content)

delete_resources

Remove resources

restore_resources

Restore soft-deleted resources

create_skill

Create markdown skill document

update_skill

Update skill metadata/references

search_skills

Find skills by tags or query

get_skills

Get skills by name with content

link_skill

Link skill to facts/resources/skills

sync_skill

Sync skill after file edit

delete_skills

Remove skills

get_dependency_graph

Get skill dependency tree

restore_skills

Restore soft-deleted skills

submit_execution_logs

Record command/test/build executions

search_execution_logs

Find executions by query, tags, success

get_execution_log

Get execution details by ID

create_tags

Create organizational tags

list_tags

List tags with usage counts

update_tags

Update tag descriptions

prune_orphan_tags

Clean up unused orphan tags

get_config

Get a configuration value by key

set_config

Set a configuration value

delete_config

Delete a configuration value

list_config

List all configuration with schema

get_config_schema

Get available options with descriptions

get_preference_prompt

Get natural language preference prompt

get_user_preferences

Get structured preference data

infer_preference

Update preference from user behavior

reset_preferences

Reset preferences to defaults

check_stale

Find stale resources and dependencies

mark_resources_refreshed

Mark resources as current

get_knowledge_context

Build context from tags (facts/resources/skills)

build_skill_context

Get skill with formatted content and refs

get_maintenance_report

Generate staleness/maintenance report

get_refresh_guide

Get instructions for refreshing a resource

get_agent_guide

Get the agent workflow guide (call first)

get_concept_guide

Get conceptual overview and design philosophy

get_config_guide

Get configuration guide with all options