MCP Server

The Persona MCP server gives any MCP-compatible agent native access to manage its identity, run pattern-based discovery, and evolve its beliefs from real behavior. No HTTP requests needed. Just call tools like add_belief and run_discovery.

Works with Claude Desktop, Claude Code, Cursor, Windsurf, and any MCP-compatible client.

Quick Setup

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "persona": {
      "command": "npx",
      "args": ["-y", "mcp-persona"],
      "env": {
        "PERSONA_API_KEY": "ln_your_key"
      }
    }
  }
}

Claude Code

claude mcp add persona -- npx -y mcp-persona

Cursor / Windsurf

Add to your MCP settings:

{
  "persona": {
    "command": "npx",
    "args": ["-y", "mcp-persona"],
    "env": {
      "PERSONA_API_KEY": "ln_your_key"
    }
  }
}

Zero-Config (No API Key)

Don't have an API key? The register tool creates an account and stores the key automatically for your session:

{
  "mcpServers": {
    "persona": {
      "command": "npx",
      "args": ["-y", "mcp-persona"]
    }
  }
}

Then just call register as your first tool. The key is stored in memory for the rest of the session.

Environment Variables

Name	Required	Description
`PERSONA_API_KEY`	No	API key (`ln_` prefix). Use `register` tool to get one automatically.
`PERSONA_BASE_URL`	No	API base URL. Default: `https://agentpersona.live/api/v1`

Tools (44)

Registration

Tool	Description
`register`	Create account + org + API key in one call. Zero fields required. Key auto-stored for session.

Agents

Tool	Description
`get_agent`	Get agent profile (name, job title, description, system prompt)
`list_agents`	List all agents in the org
`update_agent`	Update agent profile fields
`get_identity`	Get resolved identity (beliefs + responsibilities merged from org, group, and agent layers)

Beliefs

Tool	Description
`add_belief`	Add a belief. Categories: axiom, principle, voice, preference, boundary. Semantic dedup auto-checks.
`list_beliefs`	List beliefs with optional filters (category, status, starred, hidden)
`update_belief`	Update statement, category, status, starred, or hidden. Works across all scopes.
`delete_belief`	Delete or hide (PBD-discovered beliefs are hidden to preserve history)
`add_belief_bulk`	Add up to 100 beliefs at once

Responsibilities

Tool	Description
`add_responsibility`	Add a responsibility. Categories: ownership, execution, collaboration, deliverables, monitoring.
`list_responsibilities`	List with optional filters
`update_responsibility`	Update across all scopes
`delete_responsibility`	Delete or hide
`add_responsibility_bulk`	Add up to 100 at once

Organization Identity

Tool	Description
`get_org_beliefs`	Get org-level beliefs (cascade to all agents)
`add_org_belief`	Add an org-level belief
`get_org_responsibilities`	Get org-level responsibilities
`add_org_responsibility`	Add an org-level responsibility
`get_org_summary`	Full org overview: groups, agents, beliefs, responsibilities, counts

Group Identity

Tool	Description
`get_group_beliefs`	Get group-level beliefs (cascade to group members)
`add_group_belief`	Add a group-level belief
`get_group_responsibilities`	Get group-level responsibilities
`add_group_responsibility`	Add a group-level responsibility

Discovery (PBD)

Tool	Description
`run_discovery`	Trigger Pattern-Based Discovery. Extracts behavioral patterns from content, proposes new beliefs. Returns jobId.
`get_discovery_stats`	PBD quality metrics: funnel, approval rates, dedup stats, category distribution
`list_observations`	View raw observations extracted by PBD
`list_signals`	View clustered signals with n-counts and stability scores
`review_pending`	List all beliefs and responsibilities awaiting approval

Content Sources

Tool	Description
`list_sources`	List content sources for an agent (GitHub, website, Twitter, RSS)
`create_source`	Create a content source to import content for PBD
`sync_source`	Trigger sync to import new content and run PBD
`list_content`	List imported content items

Conversations

Tool	Description
`create_conversation`	Start a conversation session with agent participants
`add_message`	Add a message to a conversation
`complete_conversation`	Mark done. Auto-publishes to PBD if org setting enabled.

Evolution

Tool	Description
`snapshot_genome`	Capture beliefs + responsibilities as a point-in-time snapshot
`get_diff`	See what changed since a specific date
`run_dedup`	Scan and merge semantically duplicate beliefs. dryRun=true previews.
`run_consensus`	Detect shared beliefs across agents in a group or org

Jobs

Tool	Description
`get_job`	Check progress of a background job (PBD, sync, dedup, consensus)
`list_jobs`	List jobs with optional filters

Bootstrap

Tool	Description
`get_recommendations`	Get identity template recommendations based on agent profile
`apply_recommendations`	Batch apply selected templates

Resources (5)

Resource	URI	Description
Belief Categories	`persona://schema/belief-categories`	Category definitions with examples (axiom, principle, voice, preference, boundary)
Responsibility Categories	`persona://schema/responsibility-categories`	Category definitions (ownership, execution, collaboration, deliverables, monitoring)
About	`persona://about`	Platform overview with key concepts and links
PBD Stats	`persona://pbd/stats`	Discovery quality metrics and funnel

Prompts (3)

Prompt	Description
`get_started`	Walk through registration, creating an agent, and adding first beliefs
`evolve_identity`	Guide for connecting content sources and running PBD discovery
`review_pending`	Review and approve/reject pending beliefs and responsibilities

How It Works

The MCP server is a thin adapter. It translates MCP tool calls into REST API requests to agentpersona.live/api/v1 and returns the responses. Your agent never sees HTTP.

Agent calls add_belief("axiom", "Code is a liability")
    |
    v
MCP server sends POST /api/v1/beliefs { category: "axiom", statement: "..." }
    |
    v
Persona API checks semantic dedup, creates belief, returns result
    |
    v
Agent receives structured response with next_steps

All responses include HATEOAS next_steps arrays so agents can discover available actions from any state.

Registry Links

Registry	Link
npm	mcp-persona
GitHub	live-neon/persona-mcp

REST API

Prefer raw HTTP? See the full API reference.