Skip to content

MCP Channel

Connect your AI coding assistant to IdentityScribe and query identity data, discover schemas, and monitor system health—directly from your IDE.

IdentityScribe exposes two MCP servers:

EndpointPurposeTools
/mcpData — Schema discovery, entry search, history, documentationdescribe, search, lookup, changes, help, ref
/observe/mcpObserve — System health, status, diagnosticsobserve

Works with: Cursor, Claude Desktop, VS Code, Windsurf, Codex, Gemini CLI, Goose, and any MCP-compatible client.

Related:

Enable the MCP channels:

Terminal window
# Data MCP at /mcp
export SCRIBE_MCP_ENABLED=true
# Observe MCP at /observe/mcp (optional)
export SCRIBE_MONITORING_MCP_ENABLED=true

Then configure your AI client to connect.



Once connected, your AI assistant can query identity data and check system health. Just ask in plain English.

Explore the schema

“What entry types are in the directory?”

“Show me the user attributes”

“What fields does the Group type have?”

Find entries

“Find users with @acme.com email addresses”

“Show me all groups that contain ‘admin’ in the name”

“List the first 50 roles”

Look up a specific entry

“Get the user with ID dXNlcjo5aXg”

“Show me the entry with DN cn=john,ou=users,dc=example,dc=com”

“What did user 12345 look like yesterday?” (point-in-time lookup)

View change history

“What changed in the last 24 hours?”

“Show me the history for user dXNlcjo5aXg”

“What attributes were modified on this entry?”

Filter syntax help

If you’re building filters manually, ask for help:

“How do I write a filter to find users created after January 2024?”

“What filter syntax does IdentityScribe support?”

Look up reference docs

“What does error SCRIBE-1001 mean?”

“Tell me about the scribe.query.duration metric”

Check system health

“Is the system healthy?”

“What’s the current status?”

“Are all services running?”

Diagnose issues

“Run diagnostics”

“What’s wrong with the system?”

“Check for any problems”

View performance metrics

“Show me the error rates”

“What’s the query latency?”

“How many requests per second are we handling?”

Check configuration

“What channels are enabled?”

“Show me the current channel bindings”


The Data MCP exposes these resources for AI agents to access schemas and documentation:

URIDescription
scribe://openapiOpenAPI 3.1 specification (JSON)
scribe://graphql-schemaGraphQL schema (SDL)
scribe://ref-indexReference documentation index (JSON)

Terminal window
# Enable
SCRIBE_MCP_ENABLED=true
# Bind to specific socket(s) — optional
SCRIBE_MCP_SOCKET=admin,internal
# Query limits
SCRIBE_MCP_DEFAULT_LIMIT=25
SCRIBE_MCP_MAX_LIMIT=100
Terminal window
# Enable
SCRIBE_MONITORING_MCP_ENABLED=true
# Bind to specific socket(s) — optional
SCRIBE_MONITORING_MCP_SOCKET=admin

  1. Check IdentityScribe is running: curl http://localhost:8080/observe/health
  2. Verify MCP is enabled in config
  3. Confirm the URL in your client matches the server address
  1. Check the MCP channel is enabled (SCRIBE_MCP_ENABLED=true)
  2. Restart your AI client after config changes
  3. Try asking “What tools are available?” to verify connection

Large result sets are automatically paginated. If you need more results, ask:

“Show me the next page of results”

“Get 100 users instead of 25”

If you see HTTP 500 errors when connecting, check that HTTP/1.1 is enabled on the socket:

http.protocols.http-1-1.enabled = true

MCP uses Server-Sent Events (SSE) which requires HTTP/1.1. HTTP/2-only sockets will fail.

Ask your AI assistant directly:

“What does error SCRIBE-1001 mean?”

“Help me understand this error”

The ref tool can look up any error code, metric, or config key from the reference documentation.