Lucerna MCP

CLI or API | MCP

Documentation — Installation, deployment, and usage across the API, CLI, MCP, official documentation.

---

Table of Contents

• Overview
• Installation
• Usage
• Architecture
• Environment Variables
• MCP Tools
• Documentation

Overview

lucerna-mcp exposes a standardized interface to the platform's OpenSearch deployment (CA-50, http://localhost:9200) via the Model Context Protocol — index/alias/mapping/settings/rollover admin, BM25/k-NN/hybrid search, ingest pipeline admin, and document-level-security (DLS) policy read/apply. A trigger-only opensearch_reindex_from_kg tool records a rebuild request as an :IndexingRun node; the actual reindex is CA-24's own CDC-fed indexer.

The single hardest invariant this package holds: every request carries the calling principal's own token, attached fresh on every HTTP call — never a fixed/service-level credential. lucerna_mcp.auth.get_client() is the ONLY client factory in the package; there is no fallback path. See Architecture and AGENTS.md for how this is enforced structurally.

k-NN is disabled cluster-wide (pre-AVX2 homelab nodes — no AVX2-free OpenSearch k-NN build exists upstream). opensearch_knn_search/ opensearch_hybrid_search exist per this package's scope but raise a typed, named error rather than a bare 500 when the plugin refuses the operation.

Installation

Pick the extra that matches what the operator want to run:

Extra	Installs	Use when
lucerna-mcp[mcp]	Connector-focused MCP server (scion-sdk — FastMCP/FastAPI + epistemic-graph[full])	The operator only run the MCP server (smallest install / image)
lucerna-mcp[all]	Everything (mcp + agent + logfire)	Development / both surfaces

uv pip install "lucerna-mcp[mcp]"

Container images (:mcp)

One multi-stage docker/Dockerfile builds the MCP-only image (--target mcp):

docker build --target mcp   -t lucerna-mcp:mcp    .

Usage

Run the MCP server directly:

python -m lucerna_mcp

MCP Configuration Example (stdio)

{
  "mcpServers": {
    "lucerna-mcp": {
      "command": "uv",
      "args": ["run", "lucerna-mcp"],
      "env": {
        "MCP_TOOL_MODE": "intent",
        "OPENSEARCH_URL": "http://localhost:9200",
        "OPENSEARCHTOOL": "True",
        "INGESTTOOL": "True",
        "ENABLE_DELEGATION": "True",
        "OIDC_CONFIG_URL": "http://localhost:8080/realms/homelab/.well-known/openid-configuration",
        "OIDC_CLIENT_ID": "",
        "OIDC_CLIENT_SECRET_REF": "",
        "OPENSEARCH_KEYCLOAK_CLIENT_ID": "opensearch"
      }
    }
  }
}

Architecture

api/api_client_opensearch.py's OpenSearchApi wraps opensearchpy.OpenSearch directly (the fleet's one new dependency this package adds). Its transport auth is api/api_client_base.py's _PerCallBearerAuth — a requests.auth.AuthBase callable that opensearchpy's RequestsHttpConnection invokes fresh on EVERY outbound request (checked against opensearch-py 3.2.0's own source: RequestsHttpConnection.__init__ assigns http_auth directly to requests.Session.auth, and requests calls an AuthBase.__call__ per request, never once at session construction).

auth.py's get_client() is the ONLY client factory in the package. It exchanges the calling MCP principal's own token for one scoped to the opensearch Keycloak client audience via the fleet's shared scion_sdk.mcp.delegated_auth (RFC 8693 token exchange) — and, unlike gitlab-api/twenty-mcp's use of the same helper, it has no fallback to a fixed/service credential: if delegation isn't enabled or the exchange fails, it raises. No tool in mcp/mcp_opensearch.py or kg_ingest.py accepts a credential override parameter of any kind.

mcp/mcp_opensearch.py registers four tool groups (index, search, pipelines, security) via one register_tool_surface(...) call in mcp_server.py, matching the fleet's one-registration-call convention. kg_ingest.py registers the fifth group (opensearch_reindex_from_kg, tagged ingest/kg/kg_ingest) — a trigger-only tool that records an :IndexingRun node through the required native_ingest (Wire-First) authority; it never performs the reindex itself (CA-24 does).

opensearch_apply_dls_bundle validates a caller-supplied DEC-CA-04 policy bundle's shape (renderings.opensearch, governs: ["M1"], graph) before applying it — this package contains no code path that builds a DLS predicate from scratch.

Environment Variables

Package environment variables

Variable	Example	Description
OPENSEARCH_URL	http://localhost:9200
ENABLE_DELEGATION	True	lucerna-mcp has NO fixed/service-credential fallback — every request carries the calling principal's own token, exchanged for one scoped to the audience below. These flags are the fleet-shared MCP OIDC-delegation surface (scion_sdk.mcp.server_factory / delegated_auth), not lucerna-mcp-specific, but MUST be enabled for this package's tools to work at all:
OIDC_CONFIG_URL	http://localhost:8080/realms/homelab/.well-known/openid-configuration
OIDC_CLIENT_ID	—
OIDC_CLIENT_SECRET_REF	—
OPENSEARCH_KEYCLOAK_CLIENT_ID	opensearch	Audience/scope this package exchanges the caller's token for — defaults to OPENSEARCH_KEYCLOAK_CLIENT_ID if --audience isn't passed at MCP startup.
OPENSEARCH_TLS_PROFILE	—
OPENSEARCH_TLS_PROFILE_REF	—
OPENSEARCHTOOL	True
INGESTTOOL	True

Inherited scion-sdk variables (apply to every connector)

Variable	Example	Description
TRANSPORT	stdio	MCP transport: stdio | streamable-http | sse
HOST	127.0.0.1	Loopback bind host (set an authenticated ingress explicitly)
PORT	8000	Bind port (HTTP transports)
MCP_TOOL_MODE	intent	Tool surface: intent | condensed | verbose | both
MCP_ENABLED_TOOLS	—	Comma-separated tool allow-list
MCP_DISABLED_TOOLS	—	Comma-separated tool deny-list
MCP_ENABLED_TAGS	—	Comma-separated tag allow-list
MCP_DISABLED_TAGS	—	Comma-separated tag deny-list
EUNOMIA_TYPE	none	Authorization mode: none | embedded | remote
EUNOMIA_POLICY_FILE	mcp_policies.json	Embedded Eunomia policy file
EUNOMIA_REMOTE_URL	—	Remote Eunomia authorization server URL
ENABLE_OTEL	False	Enable OpenTelemetry export
OTEL_EXPORTER_OTLP_ENDPOINT	—	OTLP collector endpoint
MCP_CLIENT_AUTH	—	Outbound MCP child auth: oidc-client-credentials | basic | none
MCP_BASIC_AUTH_USERNAME	—	HTTP Basic username (MCP_CLIENT_AUTH=basic)
MCP_BASIC_AUTH_PASSWORD_REF	secret://identity/mcp-basic-password	Runtime secret reference for HTTP Basic auth (MCP_CLIENT_AUTH=basic)
DEBUG	False	Verbose logging
PYTHONUNBUFFERED	1	Unbuffered stdout (recommended in containers)
MCP_URL	http://localhost:8000/mcp	URL of the MCP server the agent connects to
PROVIDER	openai	LLM provider for the agent
MODEL_ID	gpt-4o	Model id for the agent

10 package + 22 inherited variable(s). Auto-generated from .env.example + the shared scion-sdk set — do not edit.

Variable	Required	Notes
OPENSEARCH_URL	recommended	Bare OpenSearch origin. Defaults to http://localhost:9200.
ENABLE_DELEGATION	✅	Must be True — this package has no other credential path.
OIDC_CONFIG_URL	✅	OIDC discovery document URL (shared scion-sdk MCP surface).
OIDC_CLIENT_ID / OIDC_CLIENT_SECRET_REF	✅	Confidential client used for the RFC 8693 token exchange.
OPENSEARCH_KEYCLOAK_CLIENT_ID	optional	Delegated-token audience. Defaults to opensearch.
OPENSEARCH_TLS_PROFILE / OPENSEARCH_TLS_PROFILE_REF	optional	Named outbound TLS trust policy.
OPENSEARCHTOOL	optional	Tool-group toggle (set False to disable the index/search/pipelines/security tools). Default True.
INGESTTOOL	optional	Tool-group toggle for opensearch_reindex_from_kg. Default True.
MCP_TOOL_MODE	optional	condensed | verbose | both | intent (inherited from scion-sdk).

MCP Tools

Tool	Group	Purpose
opensearch_create_index	index	Create an index (idempotent)
opensearch_update_mapping	index	Merge field mappings
opensearch_manage_alias	index	Add/remove/swap aliases
opensearch_get_aliases	index	Read-only alias inspection
opensearch_update_settings	index	Update dynamic index settings
opensearch_rollover	index	Roll a write alias to a new backing index
opensearch_delete_index	index	Remove an index
opensearch_search	search	BM25 lexical search (capped at 100 hits)
opensearch_knn_search	search	k-NN vector search (typed error — plugin disabled)
opensearch_hybrid_search	search	Lexical + vector hybrid
opensearch_ingest_pipeline	pipelines	Get/put an ingest pipeline
opensearch_read_dls_rules	security	Read-only: one role's DLS query
opensearch_apply_dls_bundle	security	Validate + apply a CA-16/26-rendered bundle
opensearch_reindex_from_kg	ingest	Trigger-only: records an :IndexingRun node

Documentation

See docs/ for architecture, configuration, and deployment notes, and AGENTS.md for domain-specific traps (the calling-principal invariant, the k-NN-disabled degrade path, and the DLS-bundle-never-hand-authored rule).

Available MCP Tools

Condensed action-routed tools (MCP_TOOL_MODE=condensed)

MCP Tool	Toggle Env Var	Description
opensearch_apply_dls_bundle	OPENSEARCHTOOL	Apply a CA-16/26-rendered policy bundle's OpenSearch renderings.
opensearch_ingest_pipeline	OPENSEARCHTOOL	Read or define an ingest pipeline.
opensearch_reindex_from_kg	INGESTTOOL	Request a reindex of index_pattern from the KG's current state.

Verbose 1:1 API-mapped tools (MCP_TOOL_MODE=verbose or both)

18 per-operation tools — one per public API method (click to expand)

MCP Tool	Toggle Env Var	Description
opensearch_apply_dls_bundle_rendering	OPEN_SEARCH_APITOOL	Apply ONE already-validated renderings.opensearch entry.
opensearch_cluster_health	OPEN_SEARCH_APITOOL	Invoke the cluster_health operation.
opensearch_create_index	OPENSEARCHTOOL	Invoke the create_index operation.
opensearch_delete_index	OPENSEARCHTOOL	Invoke the delete_index operation.
opensearch_get_aliases	OPENSEARCHTOOL	Invoke the get_aliases operation.
opensearch_get_ingest_pipeline	OPEN_SEARCH_APITOOL	Invoke the get_ingest_pipeline operation.
opensearch_hybrid_search	OPENSEARCHTOOL	OpenSearch's native hybrid (lexical + vector) query compound clause.
opensearch_ingest_pipeline__get	MUTATINGTOOL	Read or define an ingest pipeline.
opensearch_ingest_pipeline__put	MUTATINGTOOL	Read or define an ingest pipeline.
opensearch_knn_plugin_enabled	OPEN_SEARCH_APITOOL	Live check of the k-NN plugin's runtime-enabled flag (never assumed).
opensearch_knn_search	OPENSEARCHTOOL	Invoke the knn_search operation.
opensearch_manage_alias	OPENSEARCHTOOL	Apply one or more alias actions ({"add": {...}}/{"remove": {...}}).
opensearch_put_ingest_pipeline	OPEN_SEARCH_APITOOL	Invoke the put_ingest_pipeline operation.
opensearch_read_dls_rules	OPENSEARCHTOOL	Read one OpenSearch security-plugin role's index permissions/DLS query.
opensearch_rollover	OPENSEARCHTOOL	Invoke the rollover operation.
opensearch_search	OPENSEARCHTOOL	Invoke the search operation.
opensearch_update_mapping	OPENSEARCHTOOL	Invoke the update_mapping operation.
opensearch_update_settings	OPENSEARCHTOOL	Invoke the update_settings operation.

3 action-routed tool(s) · 18 verbose 1:1 tool(s). Each is enabled unless its <DOMAIN>TOOL toggle is set false; MCP_TOOL_MODE selects the surface (intent default — the six verb-tools, granular set loaded on demand · condensed action-routed · verbose 1:1 · both). Auto-generated — do not edit.

---

Repository Owners

---

Contribute

Contributions are welcome! Please ensure code quality by running local checks before submitting pull requests:

• Format code using ruff format .
• Lint code using ruff check .
• Validate type-safety with mypy .
• Ran test suites using pytest

Deploy with scion-sdk-deployment

Provision this package with the consolidated scion-sdk-deployment workflow. It selects an installed-package, editable-source, or immutable-container path; records only runtime secret and TLS-profile references in AgentConfig; and runs doctor, registration, policy, observability, and rollback gates. Ask your agent to "deploy lucerna-mcp with scion-sdk-deployment".

Install mode	Command
Installed package	uv tool install "lucerna-mcp[mcp]", then run lucerna-mcp
Immutable container	deploy registry.example.invalid/lucerna-mcp@sha256:<digest> through the operator-selected orchestrator

The repository embeds no deployment profile, credential value, certificate path, or environment-specific endpoint. Supply those at runtime through AgentConfig and the configured secret provider.

Governed capability contract

This package ships a compact canonical skill surface with specialist procedures kept as referenced workflows. The current MCP tools, skill metadata, connector_manifest.yml, ontology, mappings, shapes, fixtures, migrations, tool-schema fingerprints, and certification metadata form one versioned capability contract. Validate them together; do not rely on stale tool names or historical per-task skill wrappers.

Runtime endpoints, credentials, certificate trust, tenant identity, retention, and observability policy are deployment inputs and are never packaged values. See Configuration, trust, and privacy before enabling a network transport, connector ingestion, Arbora delegation, or trace export.