MCP Server Configuration

MDDB has a built-in MCP (Model Context Protocol) server implementing both the stateless 2026-07-28 specification and the handshake-based 2025-11-25, side by side. This document covers all MCP configuration options.

→ Protocol revisions: what each one requires | → LLM client setup (Claude, Cursor, ChatGPT, Ollama) | → Custom YAML tools

Transports

Transport	Endpoint	Spec Version	Status
Stdio	stdin/stdout	2026-07-28 + 2025-11-25	Default for Claude Desktop
Streamable HTTP	`POST /mcp` (plus `GET`/`DELETE` for handshake clients)	2026-07-28 + 2025-11-25	Recommended for remote
SSE (legacy)	`GET /sse` + `POST /message`	2024-11-05	Backward compatible

GET /mcp and DELETE /mcp belong to the handshake era only; 2026-07-28 removed sessions and the standalone stream. Both stay available for the clients that still use them.

All transports run on the MCP port (default: 9000, configurable via MDDB_MCP_ADDR).

Protocol Revisions

MDDB serves two revisions at once, and each request declares its own: one carrying io.modelcontextprotocol/protocolVersion in params._meta is served statelessly under 2026-07-28, and one without it is served under 2025-11-25 as before. A handshake client and a stateless client can share a port, a connection, or a stdio process.

→ Full details of both revisions — required _meta fields, server/discover, the mirrored HTTP headers, caching hints, error codes, and what 2026-07-28 removed.

initialize still negotiates for handshake clients, and only ever offers a revision that has a handshake:

Client sends in `initialize`	MDDB answers with
a handshake revision MDDB speaks	that revision
anything else (including 2026-07-28)	its own newest handshake revision, and the client decides whether to continue

GET /v1/config reports the current state without a request, which cannot be learned from a handshake — that answers for one client, and cannot show that a pin is in force or what else would have been accepted:

{
  "protocols": {
    "mcp": {
      "enabled": true,
      "addr": ":9000",
      "stdio": false,
      "revision": "2026-07-28",
      "supportedRevisions": ["2026-07-28", "2025-11-25"],
      "revisionPinned": false,
      "handshakeRevision": "2025-11-25"
    }
  }
}

Pinning a revision

MDDB_MCP_PROTOCOL_VERSION forces one revision and disables negotiation entirely. Set it when a client's own version handling is the thing that is broken; leave it unset otherwise, because a pin also stops MDDB from serving clients that ask for anything else.

MDDB_MCP_PROTOCOL_VERSION=2026-07-28

Pinning the stateless revision takes the handshake off the table: initialize is then refused with -32022 naming what the server does serve.

A pin naming a revision this build does not implement stops the server at startup, listing what it can speak. Accepting it would produce a handshake that succeeds and a session that then misbehaves, far from the cause.

The MCP-Protocol-Version header

Required on every stateless POST, where it must also match the revision in the body's _meta — a disagreement is -32020. For a handshake client the header stays optional, and naming a revision MDDB cannot speak (or one that disagrees with a pin) is refused with 400 and a body naming what would work.

curl -X POST http://localhost:9000/mcp \
  -H "MCP-Protocol-Version: 2024-01-01" -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}'
# {"error":"unsupported MCP revision 2024-01-01; this server speaks: 2026-07-28, 2025-11-25"}

Streamable HTTP (Recommended)

A stateless client (2026-07-28) sends the call it wants, with no handshake before it:

curl -X POST http://localhost:9000/mcp \
  -H "Content-Type: application/json" \
  -H "MCP-Protocol-Version: 2026-07-28" -H "Mcp-Method: tools/list" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{"_meta":{
        "io.modelcontextprotocol/protocolVersion":"2026-07-28",
        "io.modelcontextprotocol/clientCapabilities":{}}}}'

A handshake client (2025-11-25) opens as it always did:

# Initialize
curl -X POST http://localhost:9000/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'

# Session ID returned in MCP-Session-Id header — use it for all subsequent requests

Client Configuration

{
  "mcpServers": {
    "mddb": {
      "transport": "streamable-http",
      "url": "http://localhost:9000/mcp"
    }
  }
}

Environment Variables

Core

Env Var	Default	Description
`MDDB_MCP_ENABLED`	`true`	Enable MCP HTTP server
`MDDB_MCP_ADDR`	`:9000`	MCP HTTP listen address
`MDDB_MCP_PORT`	`9000`	MCP HTTP port (alternative to ADDR)
`MDDB_MCP_STDIO`	`false`	Run in MCP stdio mode (replaces HTTP/gRPC)
`MDDB_MCP_DOMAIN`	—	Server hostname for Panel config generation
`MDDB_MCP_MODE`	—	Access mode override: `read`, `write`, or `wr` (inherits `MDDB_MODE`)

Server Info

Env Var	Default	Description
`MDDB_MCP_SERVER_NAME`	`mddbd`	Server name in MCP initialize response
`MDDB_MCP_SERVER_DESCRIPTION`	—	Server description
`MDDB_MCP_SERVER_VENDOR`	—	Vendor name
`MDDB_MCP_SERVER_HOMEPAGE`	—	Homepage URL
`MDDB_MCP_INSTRUCTIONS`	—	System prompt for LLM — guidance on how to use this server

Tools

Env Var	Default	Description
`MDDB_MCP_CONFIG`	—	Path to YAML file with custom tool definitions
`MDDB_MCP_BUILTIN_TOOLS`	`true`	Set to `false` to hide all 81 built-in tools (only custom tools exposed)

API Key Authentication

Env Var	Default	Description
`MDDB_MCP_API_KEY_ENABLED`	`false`	Require API key for MCP HTTP access
`MDDB_MCP_API_KEYS`	—	Static keys: `key1:name1,key2:name2` (defined at startup)
`MDDB_MCP_API_KEY_CACHE_TTL`	`60s`	Cache TTL for dynamic key lookups

⚠️ MCP exposure vs. main auth (SEC-002). The MCP listener (default :9000) is a separate port that grants full read/write access to the database. Its own MCPAPIKeyMiddleware is inactive unless MDDB_MCP_API_KEY_ENABLED=true. To prevent MCP from becoming an unauthenticated bypass of the main API, the server reconciles MCP exposure with MDDB_AUTH_ENABLED at startup:

`MDDB_AUTH_ENABLED`	`MDDB_MCP_API_KEY_ENABLED`	MCP listener behaviour
`true`	`false`	Gated by the main AuthManager — MCP requires the same `Authorization: Bearer` / `X-API-Key` credentials as the HTTP API. Anonymous `tools/call` → `401`.
`true`	`true`	Gated by MCP API keys (as configured below).
`false`	`true`	Gated by MCP API keys.
`false`	`false`	No authentication. If MCP is bound beyond loopback (e.g. the default `:9000`), the server logs a prominent startup warning. Use only on a trusted/loopback bind, or set one of the flags above.

Two sources of keys:

Static — MDDB_MCP_API_KEYS env var (defined at startup, no restart needed to validate)
Dynamic — REST API (/v1/mcp/keys) stored in BoltDB, persists across restarts, managed without downtime

Clients authenticate via:

X-API-Key: <key> header
Authorization: Bearer <key> header
?api_key=<key> query parameter (for SSE connections)

# Static keys only
docker run -d \
  -e MDDB_MCP_API_KEY_ENABLED=true \
  -e "MDDB_MCP_API_KEYS=sk-prod:claude-prod,sk-dev:cursor-dev" \
  -p 9000:9000 \
  tradik/mddb:latest

Dynamic Key Management API

Keys stored in internal BoltDB bucket (_mcp_api_keys). Requires admin auth when MDDB_AUTH_ENABLED=true.

# Create a key (full key shown only once in response)
curl -X POST http://localhost:11023/v1/mcp/keys \
  -H "Authorization: Bearer <admin-token>" \
  -d '{"name": "claude-prod", "expiresAt": 0}'
# Response: {"key":"mcp_a1b2c3d4e5f6...","name":"claude-prod","createdAt":1711...}

# List keys (full keys NOT shown, only prefixes)
curl http://localhost:11023/v1/mcp/keys \
  -H "Authorization: Bearer <admin-token>"

# Delete a key
curl -X DELETE http://localhost:11023/v1/mcp/keys \
  -H "Authorization: Bearer <admin-token>" \
  -d '{"key": "mcp_a1b2c3d4e5f6..."}'

# Disable a key (revoke without deleting)
curl -X POST http://localhost:11023/v1/mcp/keys/disable \
  -H "Authorization: Bearer <admin-token>" \
  -d '{"key": "mcp_a1b2c3d4e5f6..."}'

Key features:

TTL support — expiresAt Unix timestamp (0 = never expires)
Disable without delete — revoke instantly, keep for audit
Cache — lookups cached for MDDB_MCP_API_KEY_CACHE_TTL (default 60s), auto-invalidated on create/delete/disable
Panel UI — manage keys from the web panel (LLM Connections → API Keys tab)

Rate Limiting

Env Var	Default	Description
`MDDB_MCP_RATE_LIMIT_ENABLED`	`false`	Enable per-client rate limiting
`MDDB_MCP_RATE_LIMIT_REQUESTS`	`100`	Max requests per window
`MDDB_MCP_RATE_LIMIT_WINDOW`	`60`	Window duration in seconds
`MDDB_MCP_RATE_LIMIT_BURST`	`20`	Burst allowance above limit
`MDDB_MCP_RATE_LIMIT_BY`	`ip`	Rate limit key: `ip`, `api_key`, or `session`

Response headers: X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset. Returns 429 Too Many Requests with Retry-After header when exceeded.

Request Logging

Env Var	Default	Description
`MDDB_MCP_LOGGING_ENABLED`	`false`	Enable structured JSON audit logs
`MDDB_MCP_LOGGING_LEVEL`	`info`	Log level: `debug`, `info`, `warn`, `error`

Logs are written to stderr in JSON format:

{"timestamp":"2026-03-26T10:30:00Z","method":"POST","path":"/mcp","status":200,"duration_ms":45,"client_ip":"1.2.3.4","key_name":"claude-prod","session_id":"abc123","user_agent":"claude-code/1.0"}

Agent Instructions

The tool schemas say what each tool accepts; they do not say which one fits a question, or that asking for a projection instead of whole documents changes the cost by more than an order of magnitude. Measured on this engine, the same full_text_search over five 12 KB documents returns ~19 700 tokens by default and ~327 with fields: ["key"].

integrations/agent-instructions/ ships that guidance in the formats agents read:

File	For
`AGENTS.md`	Paste into a project; the source every other variant is generated from
`claude-code/SKILL.md`	Claude Code skill
`cursor/mddb.mdc`	Cursor rule
`windsurf/mddb.md`	Windsurf rule

Edit AGENTS.md and run make agent-instructions; CI fails if a variant was edited directly, and a test fails if the instructions name a tool that no longer exists.

Locating a Match

full_text_search accepts highlight: true (v2.12.0+), which returns each matching fragment with the lines it occupies:

{"name": "full_text_search", "arguments": {
  "collection": "theme", "query": "background",
  "highlight": true, "fragment_size": 80, "fields": ["key"]
}}

css/style.css:158-163
"…}\n\n.hero-banner {\n  <mark>background</mark>: url(hero.png);"

fields drops the document body and highlight says where to look, so the two together answer the question without carrying the file — under 800 bytes for a 164-line stylesheet, against ~20 000 tokens for the same search returning whole documents. Projections keep the fragments deliberately: dropping them would leave a caller with neither the text nor its location.

Lower fragment_size for source. The default of 150 is a byte budget tuned for prose; code lines are short, so it covers about fifteen lines of CSS.

Built-in Tool Catalog

All 81 built-in tools, grouped by area. Tool inputs are self-describing via MCP schema discovery (tools/list); the semantic_search tool additionally supports retrieval_mode/window_size (passage-level results) and mmr/mmr_lambda (result diversification).

Documents

Tool	Description
`add_document`	Add or update a document in MDDB.
`search_documents`	Search documents with filters and sorting.
`delete_document`	Delete a document from MDDB.
`import_url`	Import a markdown document from a URL.
`set_ttl`	Set or remove time-to-live on a document.
`validate_document`	Validate document metadata against collection schema without adding the document.
`add_documents_batch`	Add multiple documents to a collection in a single batch operation.
`delete_documents_batch`	Delete multiple documents from a collection in a single batch operation.
`get_meta_keys`	List all unique metadata keys and their values for a collection.
`get_checksum`	Get a CRC32 checksum for a collection that changes when documents are modified.
`ingest_documents`	Bulk ingest documents with URL-based key derivation, YAML frontmatter extraction, content deduplication, and automatic metadata injection.
`upload_file`	Upload a file and convert it to markdown.

Search

Tool	Description
`aggregate`	Compute metadata facets (value counts) and date histograms over a collection, with optional metadata pre-filtering.
`semantic_search`	Search documents by meaning using semantic similarity.
`full_text_search`	Search documents by text content using full-text search with term matching and relevance scoring.
`autocomplete`	Prefix autocomplete — returns top-N terms starting with the query, ranked by document frequency.
`classify_document`	Zero-shot document classification.
`hybrid_search`	Combined sparse (FTS) + dense (vector) search with alpha blending or Reciprocal Rank Fusion.
`cross_search`	Search across multiple collections using a source document's embedding or a text query.
`find_duplicates`	Find duplicate and similar documents within a collection.

Geo

Tool	Description
`geo_search`	Find documents within a given radius (in meters) of a latitude/longitude point.
`geo_within`	Find documents inside an axis-aligned bounding box (minLat..maxLat × minLng..maxLng).
`geo_polygon`	Find documents whose coordinates fall inside a GeoJSON Polygon (outer ring + optional holes) or MultiPolygon (union of polygons).
`geo_stats`	Report per-collection indexed-point counts plus any loaded postcode datasets.
`geo_encode`	Convert a (lat, lng) pair into a geohash string of the requested precision (1..12, default 12).
`geo_decode`	Convert a geohash string back to the (lat, lng) centroid of its cell.

Vectors

Tool	Description
`vector_reindex`	Re-embed all documents in a collection.
`vector_stats`	Get vector/embedding statistics including provider info and per-collection embedding coverage.

FTS admin

Tool	Description
`fts_reindex`	Reindex full-text search for a collection.
`fts_languages`	List all supported FTS languages with their codes and names.
`list_synonyms`	List all synonym entries for a collection.
`add_synonym`	Add or update a synonym group for a term.
`delete_synonym`	Delete a synonym group for a term in a collection.
`list_stopwords`	List all stop words (default + custom) for a collection.
`add_stopwords`	Add custom stop words to a collection's FTS index.
`delete_stopwords`	Remove custom stop words from a collection.

Revisions

Tool	Description
`truncate_revisions`	Truncate revision history for a collection, keeping only the N most recent revisions per document.
`list_revisions`	List revision history for a document.
`restore_revision`	Restore a document to a previous revision by timestamp.

Curation

Tool	Description
`list_curation_rules`	(v2.9.14+) List curation rules that pin or hide documents for specific search queries.
`create_curation_rule`	(v2.9.14+) Create a rule that pins specific documents to fixed positions or hides others when a query matches.
`update_curation_rule`	(v2.9.14+) Replace an existing curation rule by id.
`delete_curation_rule`	(v2.9.14+) Remove a curation rule by id.

Automation

Tool	Description
`list_automation`	List all automation rules (webhooks, triggers, crons).
`create_automation`	Create a new automation rule (webhook target, search trigger, or cron schedule).
`get_automation`	Get a specific automation rule by ID.
`update_automation`	Update an existing automation rule.
`delete_automation`	Delete an automation rule by ID.
`test_automation`	Test a trigger rule by running its search and returning matches (dry run, no webhook fired).
`get_automation_logs`	List automation execution logs with optional filtering by rule ID and status.

Webhooks

Tool	Description
`register_webhook`	Register a webhook to receive HTTP callbacks when documents are added, updated, or deleted.
`list_webhooks`	List all registered webhooks.
`delete_webhook`	Delete a registered webhook by ID.

Schemas

Tool	Description
`set_schema`	Set JSON Schema for collection metadata validation.
`get_schema`	Get JSON Schema for a collection.
`delete_schema`	Delete/disable schema validation for a collection.
`list_schemas`	List all collection schemas.

Memory RAG

Tool	Description
`memory_start_session`	Start a new memory/conversation session for RAG.
`memory_add_message`	Add a message to an existing memory session.
`memory_recall`	Semantically recall relevant messages from past conversations.
`memory_summarize`	Generate a summary of a conversation session.
`memory_list_sessions`	List memory/conversation sessions with optional filtering by user, scenario, and sorting.
`memory_session_history`	Get the full message history of a specific conversation session, ordered chronologically.

WordPress

Tool	Description
`wordpress_publish`	(v2.11.0+) Create or update a WordPress post/page via the mddb-sync plugin, including tags, categories, custom taxonomies, meta fields and Polylang/WPML language assignment.
`wordpress_set_status`	(v2.11.0+) Change the publishing status of a WordPress post/page (publish, draft, pending, private, future, trash) via the mddb-sync plugin.

Bulk ingest

Tool	Description
`bulk_ingest_submit`	Queue a long-running bulk ingest job and return immediately with a job identifier.
`bulk_ingest_status`	Return the current status record (counters, timestamps, up to 50 errors) for a previously-submitted bulk ingest job.
`bulk_ingest_list`	List bulk ingest jobs newest-first, optionally filtered by collection.
`bulk_ingest_cancel`	Cancel a pending bulk ingest job.

Ops

Tool	Description
`get_stats`	Get MDDB server statistics.
`export_documents`	Export documents from a collection in NDJSON format.
`create_backup`	Create a backup of the MDDB database.
`restore_backup`	Restore the MDDB database from a backup file.
`update_document`	Partially update a document.
`get_document_meta`	Get document metadata without content.
`code_graph`	Code connection graph: what a document depends on and what depends on it (which stylesheet declares a selector, which pages load a script).
`delete_collection`	Delete an entire collection and all its documents, revisions, and metadata indices.
`get_collection_config`	Get configuration attributes for a collection (type, description, icon, color, custom metadata).
`set_collection_config`	Set or update configuration attributes for a collection.
`list_collection_configs`	List all collections that have custom configuration set.

Access Modes

Each protocol can have its own read/write mode:

Env Var	Protocol	Default
`MDDB_MODE`	Global	`wr`
`MDDB_API_MODE`	HTTP/REST + GraphQL	inherits global
`MDDB_GRPC_MODE`	gRPC	inherits global
`MDDB_MCP_MODE`	MCP (all transports)	inherits global
`MDDB_HTTP3_MODE`	HTTP/3 (QUIC)	inherits global

In read-only mode:

Built-in tools with readOnlyHint=true (search, stats, list, export) → allowed
Built-in tools with writes/deletes → blocked with error
Custom tools with read-only actions (semantic_search, search_documents, full_text_search) → allowed
Custom tools with write actions → blocked

# Public MCP: read-only, custom tools only, API keys, rate limited
docker run -d \
  -e MDDB_MCP_MODE=read \
  -e MDDB_MCP_BUILTIN_TOOLS=false \
  -e MDDB_MCP_API_KEY_ENABLED=true \
  -e "MDDB_MCP_API_KEYS=sk-public:external" \
  -e MDDB_MCP_RATE_LIMIT_ENABLED=true \
  -e MDDB_MCP_RATE_LIMIT_REQUESTS=60 \
  -e MDDB_MCP_LOGGING_ENABLED=true \
  -e MDDB_MCP_CONFIG=/app/tools.yml \
  tradik/mddb:latest

Protocol Features

Feature	Description
Tool Annotations	`readOnlyHint`, `destructiveHint`, `idempotentHint`, `openWorldHint` on all tools
Structured Output	`outputSchema` on 9 key tools for client-side validation
5 Prompts	`analyze-collection`, `search-help`, `summarize-collection`, `import-guide`, `rag-pipeline`
Completion	Autocomplete for collection names and prompt arguments
Logging	RFC 5424 levels (debug → emergency): `logging/setLevel` for handshake clients, `io.modelcontextprotocol/logLevel` per request for stateless ones
Progress Tokens	`notifications/progress` for long-running operations
Cursor Pagination	`tools/list` and `resources/list`
Resource Templates	`resources/templates/list` — the patterns (`mddb://{collection}/{key}`) live here rather than among the readable resources
Discovery (2026-07-28)	`server/discover` — supported revisions, capabilities and identity in one call
Caching Hints (2026-07-28)	`ttlMs` + `cacheScope` on every list and read
Notifications	`notifications/initialized`, `notifications/cancelled`
Memory RAG Tools	6 tools for conversational memory: start session, add message, recall, summarize, list sessions, history
Async Bulk Ingest Tools (v2.9.12+)	4 tools for long-running ingest jobs: submit, status, list, cancel
Autocomplete Tool (v2.9.12+)	Prefix autocomplete over the FTS index — top-N suggestions ranked by document frequency
Per-Query Boost (v2.9.12+)	`full_text_search` and `hybrid_search` tools accept a `boost` map keyed by `"metaKey:metaValue"` to boost/demote matching documents at query time

Async Bulk Ingest Tools (v2.9.12+)

Queue long-running bulk ingest jobs from LLMs without blocking on the tool call. The worker drains jobs FIFO in 500-document chunks; status records persist across restarts.

Tool	Description	Write?
`bulk_ingest_submit`	Queue a new job with documents + optional `callback_url`; returns job ID	Yes
`bulk_ingest_status`	Return current status record (counters, timestamps, errors)	No
`bulk_ingest_list`	List all jobs newest-first, optionally filtered by collection	No
`bulk_ingest_cancel`	Cancel a pending job (in-flight jobs run to completion)	Yes

Autocomplete Tool (v2.9.12+)

Tool	Description	Write?
`autocomplete`	Top-N prefix suggestions over the FTS inverted index, ranked by doc frequency; supports field scoping	No

WordPress Publishing Tools (v2.11.0+)

Publish straight from MCP into WordPress sites running the mddb-sync plugin with Remote publishing enabled. Posts and pages, tags/categories/custom taxonomies, post meta ("metafields") and Polylang/WPML language assignment + translation linking are all supported.

Tool	Description	Write?
`wordpress_publish`	Create or update a post/page (upsert by `post_id`, else `post_type`+`slug`) with content (Markdown or HTML), status, tags, categories, taxonomies, meta and language	Yes
`wordpress_set_status`	Change publishing status: `publish`, `draft`, `pending`, `private`, `future` (with `date`) or `trash`	Yes

The target site comes from the collection's config — one collection per WordPress site, matching how the sync plugin pushes content in:

// 1. Pin the WordPress target to the collection (once)
{"name": "set_collection_config", "arguments": {
  "collection": "example_com",
  "wordpress": {"url": "https://example.com", "api_key": "<publish key from Settings → MDDB Sync>"}
}}

// 2. Publish a post with tags, meta and a language
{"name": "wordpress_publish", "arguments": {
  "collection": "example_com",
  "post_type": "post",
  "title": "Hello from MDDB",
  "content_markdown": "# Hello\n\nPublished via MCP.",
  "status": "publish",
  "tags": ["mddb", "mcp"],
  "categories": ["News"],
  "meta": {"seoTitle": "Hello from MDDB"},
  "lang": "en_US"
}}

// 3. Later: unpublish it
{"name": "wordpress_set_status", "arguments": {"collection": "example_com", "post_id": 123, "status": "draft"}}

The site URL must be https:// (plain http:// only for localhost). Both tools also accept explicit site_url + api_key arguments to skip the collection config. Translations: pass lang plus translation_of: <post id> to link a new post as a translation via Polylang or WPML.

Projection note (v2.11.0, GO-019): for search_documents / full_text_search / semantic_search / hybrid_search, passing fields now also drops the document body by default — pass include_content: true explicitly to keep it.

Memory RAG Tools

Built-in MCP tools for conversational memory and RAG:

Tool	Description	Write?
`memory_start_session`	Start a new conversation session	Yes
`memory_add_message`	Add a message (auto-embedded)	Yes
`memory_recall`	Semantic/hybrid/keyword recall across sessions	No
`memory_summarize`	Generate and store session summary	Yes
`memory_list_sessions`	List sessions with filters	No
`memory_session_history`	Get chronological message history	No

Example: Agent with Memory

// 1. Start session
{"name": "memory_start_session", "arguments": {"user_id": "user-1", "scenario": "support"}}

// 2. Store messages as conversation progresses
{"name": "memory_add_message", "arguments": {"session_id": "abc...", "role": "user", "content": "How do I use vector search?"}}
{"name": "memory_add_message", "arguments": {"session_id": "abc...", "role": "assistant", "content": "Use POST /v1/vector-search..."}}

// 3. In a future session, recall relevant context
{"name": "memory_recall", "arguments": {"query": "vector search usage", "user_id": "user-1", "top_k": 5, "include_content": true}}

// 4. Summarize when done
{"name": "memory_summarize", "arguments": {"session_id": "abc..."}}

Built-in Prompts

Prompt	Arguments	Description
`analyze-collection`	`collection` (required)	Document count, metadata keys, content patterns, suggestions
`search-help`	`use_case` (required)	Which search method to use for your use case
`summarize-collection`	`collection` (required), `limit`	Topic overview, themes, distribution
`import-guide`	`source` (required)	Step-by-step import from wordpress/url/file/api/scraping
`rag-pipeline`	`collection` (required), `model`	RAG pipeline design with embedding and search strategy

← Back to README | → LLM Connections | → Custom Tools