{"id":31055433,"url":"https://github.com/knuckles-team/vector-mcp","last_synced_at":"2026-08-28T04:53:45.760Z","repository":{"id":313144298,"uuid":"1047482583","full_name":"Knuckles-Team/vector-mcp","owner":"Knuckles-Team","description":"Vector MCP Server for AI Agents - Supports ChromaDB, Couchbase, MongoDB, Qdrant, and PGVector","archived":false,"fork":false,"pushed_at":"2026-08-07T14:57:18.000Z","size":2403,"stargazers_count":15,"open_issues_count":0,"forks_count":3,"subscribers_count":1,"default_branch":"main","last_synced_at":"2026-08-09T14:41:41.589Z","etag":null,"topics":["a2a","a2a-server","ag-ui","chromadb","couchbase","mcp-server","mongodb","mongodb-atlas","pgvector","python","qdrant","qdrant-vector-database","rag","retrieval-augmented-generation"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Knuckles-Team.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2025-08-30T14:20:49.000Z","updated_at":"2026-08-07T14:58:52.000Z","dependencies_parsed_at":"2026-04-17T09:01:49.313Z","dependency_job_id":null,"html_url":"https://github.com/Knuckles-Team/vector-mcp","commit_stats":null,"previous_names":["knuckles-team/vector-mcp"],"tags_count":147,"template":false,"template_full_name":null,"purl":"pkg:github/Knuckles-Team/vector-mcp","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Knuckles-Team%2Fvector-mcp","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Knuckles-Team%2Fvector-mcp/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Knuckles-Team%2Fvector-mcp/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Knuckles-Team%2Fvector-mcp/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Knuckles-Team","download_url":"https://codeload.github.com/Knuckles-Team/vector-mcp/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Knuckles-Team%2Fvector-mcp/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":36949460,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-08-22T15:14:58.755Z","status":"online","status_checked_at":"2026-08-28T02:00:06.244Z","response_time":114,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["a2a","a2a-server","ag-ui","chromadb","couchbase","mcp-server","mongodb","mongodb-atlas","pgvector","python","qdrant","qdrant-vector-database","rag","retrieval-augmented-generation"],"created_at":"2025-09-15T04:48:05.959Z","updated_at":"2026-08-28T04:53:45.754Z","avatar_url":"https://github.com/Knuckles-Team.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# vector-mcp\n\nAction-routed MCP and agent interfaces for governed vector collection management and retrieval.\nThe native default is epistemic-graph. Secure opt-in providers cover PostgreSQL/pgvector,\nQdrant, and MongoDB Atlas.\n\n*Version: 3.1.0*\n\n\u003c!-- GOVERNED-CAPABILITY:START --\u003e\n## Governed capability\n\n- MCP tools: `vector_collection_management` and `vector_search`\n- Skill provider: the consolidated `vector-mcp-operations` workflow\n- Ontology provider: the packaged vector retrieval ontology\n- Source connector provider: a read-only vector collection inventory preset\n- Runtime configuration: AgentConfig, environment variables, and secret references\n- Privacy posture: no checked-in endpoints, credentials, personal identity, or host paths\n\u003c!-- GOVERNED-CAPABILITY:END --\u003e\n\n## Install\n\nUse the smallest extra set required by the deployment:\n\n```bash\nuvx --from 'vector-mcp[mcp]' vector-mcp\n```\n\nThe runtime requires `agent-utilities\u003e=2.0.0` and its self-contained full\nepistemic-graph engine contract. A bare numeric-only or partial engine profile is not a\nsupported deployment.\n\nFor a selected storage provider:\n\n```bash\nuv add 'vector-mcp[postgres]'\nuv add 'vector-mcp[qdrant]'\nuv add 'vector-mcp[mongodb]'\n```\n\nThe `all` extra enables every supported optional provider plus the agent, Langfuse, and\nLogfire runtimes. Production images should install only the providers they operate.\n\n## MCP configuration\n\nThe package includes a neutral agent-launch configuration containing only the command,\ncondensed tool mode, and tool toggles. Runtime values are inherited from AgentConfig or\ninjected by the operator. Detailed instructions on how to use the underlying API wrappers,\nextended schema bindings, and developer SDK references are maintained in\n[docs/index.md](docs/index.md).\n\n---\n\n## MCP\n\nThis server utilizes dynamic Action-Routed tools to optimize token overhead and maximize IDE compatibility.\n\n### Available MCP Tools\n\n_Auto-generated from the live MCP server — do not edit by hand._\n\n\u003c!-- MCP-TOOLS-TABLE:START --\u003e\n\n#### Condensed action-routed tools (`MCP_TOOL_MODE=condensed`)\n\n| MCP Tool | Toggle Env Var | Description |\n|----------|----------------|-------------|\n| `vector_collection_management` | `COLLECTION_MANAGEMENTTOOL` | Manage collection management operations. |\n| `vector_search` | `SEARCHTOOL` | Manage search operations. |\n\n_2 action-routed tool(s) · 0 verbose 1:1 tool(s). Each is enabled unless its `\u003cDOMAIN\u003eTOOL` 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._\n\u003c!-- MCP-TOOLS-TABLE:END --\u003e\n\nDetailed tool schemas, parameter shapes, and validation constraints are preserved in [the usage guide](docs/usage.md).\n\n### Dynamic Tool Selection \u0026 Visibility\n\nThis MCP server supports dynamic toolset selection and visibility filtering at runtime. This allows you to restrict the set of exposed tools in order to prevent blowing up the LLM's context window.\n\nYou can configure tool filtering via multiple input channels:\n\n- **CLI Arguments:** Pass `--tools` or `--toolsets` (or their disabled counterparts `--disabled-tools` and `--disabled-toolsets`) during startup.\n- **Environment Variables:** Define standard environment variables:\n  - `MCP_ENABLED_TOOLS` / `MCP_DISABLED_TOOLS`\n  - `MCP_ENABLED_TAGS` / `MCP_DISABLED_TAGS`\n- **HTTP SSE Request Headers:** Pass custom headers during transport initialization:\n  - `x-mcp-enabled-tools` / `x-mcp-disabled-tools`\n  - `x-mcp-enabled-tags` / `x-mcp-disabled-tags`\n- **HTTP SSE Request Query Parameters:** Append query parameters directly to your transport connection URL:\n  - `?tools=tool1,tool2`\n  - `?tags=tag1`\n\nWhen query strings or parameters are supplied, an LLM-free **Knowledge Graph resolution layer** (using `DynamicToolOrchestrator`) matches query intents against known tool tags, names, or descriptions, with safe fallback and automated 24-hour background cache refreshing.\n\n---\n\n### MCP Configuration Examples\n\n\u003c!-- MCP-CONFIG-EXAMPLES:START --\u003e\n\n\u003e **Install the connector-focused `[mcp]` extra.** Examples use `vector-mcp[mcp]` to add\n\u003e FastMCP / FastAPI through `agent-utilities[mcp]`; the required Agent Utilities core\n\u003e still carries `epistemic-graph[full]`. The `[agent-runtime]` extra additionally\n\u003e enables model orchestration.\n\n#### stdio Transport (local IDEs — Cursor, Claude Desktop, VS Code)\n\n```json\n{\n  \"mcpServers\": {\n    \"vector-mcp\": {\n      \"command\": \"uvx\",\n      \"args\": [\n        \"--from\",\n        \"vector-mcp[mcp]\",\n        \"vector-mcp\"\n      ],\n      \"env\": {\n        \"MCP_TOOL_MODE\": \"intent\",\n        \"COLLECTION_MANAGEMENTTOOL\": \"True\",\n        \"DATABASE_TYPE\": \"epistemic_graph\",\n        \"LLM_SSL_VERIFY\": \"False\",\n        \"SEARCHTOOL\": \"True\",\n        \"VECTOR_DB_TYPE\": \"epistemic_graph\"\n      }\n    }\n  }\n}\n```\n\nRuntime references require an alias-aware launcher such as GraphOS. Other\nlaunchers must omit those entries and inject the resolved values through their\nown runtime secret boundary.\n\n#### Streamable-HTTP Transport (networked / production)\n\n```json\n{\n  \"mcpServers\": {\n    \"vector-mcp\": {\n      \"command\": \"uvx\",\n      \"args\": [\n        \"--from\",\n        \"vector-mcp[mcp]\",\n        \"vector-mcp\",\n        \"--transport\",\n        \"streamable-http\",\n        \"--port\",\n        \"8000\"\n      ],\n      \"env\": {\n        \"TRANSPORT\": \"streamable-http\",\n        \"HOST\": \"127.0.0.1\",\n        \"PORT\": \"8000\",\n        \"MCP_TOOL_MODE\": \"intent\",\n        \"COLLECTION_MANAGEMENTTOOL\": \"True\",\n        \"DATABASE_TYPE\": \"epistemic_graph\",\n        \"LLM_SSL_VERIFY\": \"False\",\n        \"SEARCHTOOL\": \"True\",\n        \"VECTOR_DB_TYPE\": \"epistemic_graph\"\n      }\n    }\n  }\n}\n```\n\nAlternatively, connect to a pre-deployed Streamable-HTTP instance by `url`:\n\n```json\n{\n  \"mcpServers\": {\n    \"vector-mcp\": {\n      \"url\": \"http://localhost:8000/vector-mcp/mcp\"\n    }\n  }\n}\n```\n\nRun a reviewed container image as a least-privilege stdio child (no\nlistener or published port):\n\n```bash\ndocker run -i --rm \\\n  --read-only \\\n  --cap-drop=ALL \\\n  --security-opt=no-new-privileges \\\n  --pids-limit=256 \\\n  --tmpfs /tmp:rw,noexec,nosuid,nodev,size=64m \\\n  -e TRANSPORT=stdio \\\n  -e MCP_TOOL_MODE=intent \\\n  -e COLLECTION_MANAGEMENTTOOL=True \\\n  -e DATABASE_TYPE=epistemic_graph \\\n  -e LLM_SSL_VERIFY=False \\\n  -e SEARCHTOOL=True \\\n  -e VECTOR_DB_TYPE=epistemic_graph \\\n  registry.example.invalid/vector-mcp@sha256:\u003cdigest\u003e vector-mcp\n```\n\nFor containerized network HTTP, supply an authenticated TLS ingress (or\ndirect server TLS), exact `MCP_ALLOWED_HOSTS`, and an exact trusted-proxy\nCIDR policy through the operator-owned deployment profile. The generator\ndoes not emit an unauthenticated non-loopback listener.\n\n_Auto-generated from the code-read env surface (`MCP_TOOL_MODE` + package vars) — do not edit._\n\u003c!-- MCP-CONFIG-EXAMPLES:END --\u003e\n\n\u003c!-- BEGIN GENERATED: additional-deployment-options --\u003e\n### Additional Deployment Options\n\n`vector-mcp` can also run as a **local container** (Docker / Podman / `uv`) or be\nconsumed from a **remote deployment**. The\n[Deployment guide](https://knuckles-team.github.io/vector-mcp/deployment/) has full, copy-paste\n`mcp_config.json` for all four transports — **stdio**, **streamable-http**,\n**local container / uv**, and **remote URL**:\n\n- **Local container / uv** — launch the server from `mcp_config.json` via `uvx`,\n  `docker run`, or `podman run`, or point at a local streamable-http container by `url`.\n- **Remote URL** — connect to a server deployed behind Caddy at\n  `https://vector-mcp.example.invalid/mcp` using the `\"url\"` key.\n\u003c!-- END GENERATED: additional-deployment-options --\u003e\n\n---\n\n## Environment Variables\n\n\u003c!-- ENV-VARS-TABLE:START --\u003e\n\n#### Package environment variables\n\n| Variable | Example | Description |\n|----------|---------|-------------|\n| `HOST` | `127.0.0.1` |  |\n| `PORT` | `8000` |  |\n| `TRANSPORT` | `stdio` | options: stdio, streamable-http, sse |\n| `ENABLE_OTEL` | — |  |\n| `EMBEDDING_TLS_PROFILE_REF` | `secret://runtime/embedding-tls-profile` | Configure AgentConfig EMBEDDING_MODELS and its referenced runtime credentials. |\n| `LLM_BASE_URL` | `http://localhost:8000/v1` | embedding/LLM API base url |\n| `LLM_TOKEN` | secret-injected | bearer token for the embedding/LLM endpoint |\n| `LLM_API_KEY` | secret-injected | alias accepted if LLM_TOKEN is unset |\n| `LLM_SSL_VERIFY` | `False` | verify TLS for the embedding/LLM endpoint |\n| `DOCUMENT_DIRECTORY` | — | Required only for filesystem ingestion. Supply the operator-owned root at runtime. |\n| `DATABASE_TYPE` | `epistemic_graph` | Backend used when db_type is unspecified. Default is the native epistemic-graph engine (local, zero-infra, durable). Options: epistemic_graph, postgres, mongodb, qdrant. DATABASE_TYPE is the canonical variable; VECTOR_DB_TYPE is accepted as an alias for backward compatibility. |\n| `VECTOR_DB_TYPE` | `epistemic_graph` |  |\n| `DB_HOST` | — | postgres/qdrant host |\n| `DBNAME` | — | postgres/mongodb database name |\n| `DB_PORT` | `5432` |  |\n| `DB_USERNAME_REF` | `secret://runtime/db-username` |  |\n| `DB_PASSWORD_REF` | `secret://runtime/db-password` |  |\n| `MONGODB_URI_REF` | `secret://runtime/mongodb-uri` |  |\n| `QDRANT_API_KEY_REF` | `secret://runtime/qdrant-api-key` |  |\n| `QDRANT_HTTP_ALLOWED_PRIVATE_HOSTS` | — | comma-separated SSRF allowlist for a private Qdrant host |\n| `COLLECTION_MANAGEMENTTOOL` | `True` |  |\n| `SEARCHTOOL` | `True` |  |\n| `TEST_POSTGRES_CONNECTION_STRING` | `postgresql://postgres:password@localhost:5432/vectordb` |  |\n| `TEST_MONGODB_HOST` | `localhost` |  |\n| `TEST_MONGODB_PORT` | `27017` |  |\n| `TEST_MONGODB_DB` | `vectordb` |  |\n| `TEST_QDRANT_LOCATION` | `http://localhost:6333` |  |\n| `TEST_COUCHBASE_CONNECTION` | `couchbase://localhost` |  |\n| `TEST_COUCHBASE_USER` | `Administrator` |  |\n| `TEST_COUCHBASE_PASSWORD` | secret-injected |  |\n| `TEST_COUCHBASE_DB` | `vector_db` |  |\n\n#### Inherited agent-utilities variables (apply to every connector)\n\n| Variable | Example | Description |\n|----------|---------|-------------|\n| `MCP_TOOL_MODE` | `intent` | Tool surface: `intent` \\| `condensed` \\| `verbose` \\| `both` |\n| `MCP_ENABLED_TOOLS` | — | Comma-separated tool allow-list |\n| `MCP_DISABLED_TOOLS` | — | Comma-separated tool deny-list |\n| `MCP_ENABLED_TAGS` | — | Comma-separated tag allow-list |\n| `MCP_DISABLED_TAGS` | — | Comma-separated tag deny-list |\n| `EUNOMIA_TYPE` | `none` | Authorization mode: `none` \\| `embedded` \\| `remote` |\n| `EUNOMIA_POLICY_FILE` | `mcp_policies.json` | Embedded Eunomia policy file |\n| `EUNOMIA_REMOTE_URL` | — | Remote Eunomia authorization server URL |\n| `OTEL_EXPORTER_OTLP_ENDPOINT` | — | OTLP collector endpoint |\n| `MCP_CLIENT_AUTH` | — | Outbound MCP child auth: `oidc-client-credentials` \\| `basic` \\| `none` |\n| `OIDC_CLIENT_ID` | — | OIDC client id (service-account auth) |\n| `OIDC_CLIENT_SECRET_REF` | `secret://identity/oidc-client-secret` | Runtime secret reference for the OIDC service account |\n| `MCP_BASIC_AUTH_USERNAME` | — | HTTP Basic username (`MCP_CLIENT_AUTH=basic`) |\n| `MCP_BASIC_AUTH_PASSWORD_REF` | `secret://identity/mcp-basic-password` | Runtime secret reference for HTTP Basic auth (`MCP_CLIENT_AUTH=basic`) |\n| `DEBUG` | `False` | Verbose logging |\n| `PYTHONUNBUFFERED` | `1` | Unbuffered stdout (recommended in containers) |\n| `MCP_URL` | `http://localhost:8000/mcp` | URL of the MCP server the agent connects to |\n| `PROVIDER` | `openai` | LLM provider for the agent |\n| `MODEL_ID` | `gpt-4o` | Model id for the agent |\n| `ENABLE_WEB_UI` | `True` | Serve the AG-UI web interface |\n\n_31 package + 20 inherited variable(s). Auto-generated from `.env.example` + the shared agent-utilities set — do not edit._\n\u003c!-- ENV-VARS-TABLE:END --\u003e\n\n\nEvery variable the server reads, grouped by purpose. See [`.env.example`](.env.example) for the\ncanonical, copy-paste list — including the `DATABASE_TYPE` / `GRAPH_SERVICE_SOCKET` /\n`GRAPH_SERVICE_AUTH_SECRET` connection settings for the native epistemic-graph backend. Backend\nendpoints, database locations, and credentials for opt-in providers (Postgres/Qdrant/Mongo/\nChroma/Couchbase) are never README-documented literal values or MCP tool arguments — they resolve\nthrough AgentConfig and `secret://`/`env://`/`vault://` references at runtime.\n\n### MCP server / transport\n| Variable | Description | Default |\n|----------|-------------|---------|\n| `TRANSPORT` | `stdio`, `streamable-http`, or `sse` | `stdio` |\n| `HOST` | Bind host (HTTP transports) | `0.0.0.0` |\n| `PORT` | Bind port (HTTP transports) | `8000` |\n| `MCP_TOOL_MODE` | Tool surface: `condensed`, `verbose`, or `both` | `condensed` |\n| `MCP_ENABLED_TOOLS` / `MCP_DISABLED_TOOLS` | Comma-separated tool allow/deny list | — |\n| `MCP_ENABLED_TAGS` / `MCP_DISABLED_TAGS` | Comma-separated tag allow/deny list | — |\n| `PYTHONUNBUFFERED` | Unbuffered stdout (recommended in containers) | `1` |\n\n### Tool toggles\nEach action-routed tool can be disabled individually via its toggle env var (set to `false`).\nThe full list is in the [Available MCP Tools](#available-mcp-tools) table above.\n\n| Variable | Description | Default |\n|----------|-------------|---------|\n| `COLLECTION_MANAGEMENTTOOL` | Enable the collection-management tool | `True` |\n| `SEARCHTOOL` | Enable the search tool | `True` |\n\n### Telemetry \u0026 governance\n| Variable | Description | Default |\n|----------|-------------|---------|\n| `ENABLE_OTEL` | Enable OpenTelemetry export | `True` |\n| `OTEL_EXPORTER_OTLP_ENDPOINT` | OTLP collector endpoint | — |\n| `OTEL_EXPORTER_OTLP_PUBLIC_KEY` / `OTEL_EXPORTER_OTLP_SECRET_KEY` | OTLP auth keys | — |\n| `OTEL_EXPORTER_OTLP_PROTOCOL` | OTLP protocol (e.g. `http/protobuf`) | — |\n| `EUNOMIA_TYPE` | Authorization mode: `none`, `embedded`, `remote` | `none` |\n| `EUNOMIA_POLICY_FILE` | Embedded policy file | `mcp_policies.json` |\n| `EUNOMIA_REMOTE_URL` | Remote Eunomia server URL | — |\n\n### Agent CLI (full `[agent]` runtime only)\n| Variable | Description | Default |\n|----------|-------------|---------|\n| `MCP_URL` | URL of the MCP server the agent connects to | `http://localhost:8000/mcp` |\n| `PROVIDER` | LLM provider (e.g. `openai`) | `openai` |\n| `MODEL_ID` | Model id (e.g. `gpt-4o`) | `gpt-4o` |\n| `ENABLE_WEB_UI` | Serve the AG-UI web interface | `True` |\n\nSee [`.env.example`](.env.example) for a copy-paste starting point.\n\n## Provider and ontology integration\n\nThe package contributes its skills, prompts, ontology, and source connector through Python entry\npoints. The collection-inventory connector is intentionally read-only and registers collection\nmetadata, not document or embedding payloads.\n\nGenerated connector signatures must be recreated only after the installed MCP schema is observed\nand a release signing key is provided at runtime. A signature from an older tool schema or\nontology must never be copied forward.\n\n## Development checks\n\nLow-cost checks that do not launch providers:\n\n```bash\npython scripts/security_sanitizer.py\npython scripts/security_contract.py --contract .security/security-contract.json validate\npython -m compileall -q vector_mcp\n```\n\nProvider tests use mocked SDK boundaries and make no network calls. Live qualification is a\nseparate deployment gate and must use operator-supplied AgentConfig and secrets.\n\n## Documentation\n\n- [Installation](docs/installation.md)\n- [Configuration and privacy](docs/configuration.md)\n- [Deployment](docs/deployment.md)\n- [Usage](docs/usage.md)\n- [Architecture overview](docs/overview.md)\n\nThe slim `:mcp` streamable-http container (`docker/mcp.compose.yml`) publishes `:8000` with a\n`/health` check; see [Deployment](docs/deployment.md) for the full compose service definition.\n\n## License\n\nSee [LICENSE](LICENSE).\n\n\n\u003c!-- BEGIN agent-utilities-deployment (generated; do not edit between markers) --\u003e\n\n## Deploy with `agent-utilities-deployment`\n\nProvision this package with the consolidated **`agent-utilities-deployment`**\nworkflow. It selects an installed-package, editable-source, or immutable-container\npath; records only runtime secret and TLS-profile references in `AgentConfig`; and\nruns doctor, registration, policy, observability, and rollback gates. Ask your agent\nto **\"deploy `vector-mcp` with agent-utilities-deployment\"**.\n\n| Install mode | Command |\n|------|---------|\n| Installed package | `uv tool install \"vector-mcp[mcp]\"`, then run `vector-mcp` |\n| Editable source | `uv pip install -e \".[agent]\"`, then run `vector-mcp` |\n| Immutable container | deploy `registry.example.invalid/vector-mcp@sha256:\u003cdigest\u003e` through the operator-selected orchestrator |\n\nThe repository embeds no deployment profile, credential value, certificate path, or\nenvironment-specific endpoint. Supply those at runtime through `AgentConfig` and the\nconfigured secret provider.\n\n\u003c!-- END agent-utilities-deployment --\u003e\n\n---\n\n## Installation\n\nPick the extra that matches what you want to run:\n\n| Extra | Installs | Use when |\n|-------|----------|----------|\n| `vector-mcp[mcp]` | Slim MCP server only (`agent-utilities[mcp]` — FastMCP/FastAPI) | You only run the **MCP server** (smallest install / image) |\n| `vector-mcp[agent]` | Full agent runtime (`agent-utilities[agent,logfire]` — Pydantic AI + the epistemic-graph engine) | You run the **integrated agent** |\n| `vector-mcp[all]` | Everything (`mcp` + all vector backends + `agent`) | Development / both surfaces |\n\n```bash\n# MCP server only (recommended for tool hosting — slim deps)\nuv pip install \"vector-mcp[mcp]\"\n\n# Full agent runtime (Pydantic AI + epistemic-graph engine)\nuv pip install \"vector-mcp[agent]\"\n\n# Everything (development)\nuv pip install \"vector-mcp[all]\"      # or: python -m pip install \"vector-mcp[all]\"\n```\n\n### Container images (`:mcp` vs `:agent`)\n\nOne multi-stage `docker/Dockerfile` builds two right-sized images, selected by `--target`:\n\n| Image tag | Build target | Contents | Entrypoint |\n|-----------|--------------|----------|------------|\n| `knucklessg1/vector-mcp:mcp` | `--target mcp` | `vector-mcp[mcp]` — **slim**, no engine/`pydantic-ai`/`dspy`/`llama-index`/`tree-sitter` | `vector-mcp` |\n| `knucklessg1/vector-mcp:latest` | `--target agent` (default) | `vector-mcp[agent]` — **full** agent runtime + epistemic-graph engine | `vector-agent` |\n\n```bash\ndocker build --target mcp   -t knucklessg1/vector-mcp:mcp    docker/   # slim MCP server\ndocker build --target agent -t knucklessg1/vector-mcp:latest docker/   # full agent\n```\n\n`docker/mcp.compose.yml` runs the slim `:mcp` server; `docker/agent.compose.yml` runs the\nagent (`:latest`) with a co-located `:mcp` sidecar.\n\n### Knowledge-graph database (`epistemic-graph`)\n\nThe **full agent** (`[agent]` / `:latest`) embeds the **epistemic-graph** engine (pulled in\ntransitively via `agent-utilities[agent]`). For production — or to share one knowledge graph\nacross multiple agents — run **epistemic-graph as its own database container** and point the\nagent at it instead of embedding it. Deployment recipes (single-node + Raft HA), connection\nconfig, and the full database architecture (with diagrams) are documented in the\n[epistemic-graph deployment guide](https://knuckles-team.github.io/epistemic-graph/deployment/).\nThe slim `[mcp]` server does **not** require the database.\n\n---\n\n## Repository Owners\n\n\u003cimg width=\"100%\" height=\"180em\" src=\"https://github-readme-stats.vercel.app/api?username=Knucklessg1\u0026show_icons=true\u0026hide_border=true\u0026\u0026count_private=true\u0026include_all_commits=true\" /\u003e\n\n![GitHub followers](https://img.shields.io/github/followers/Knucklessg1)\n![GitHub User's stars](https://img.shields.io/github/stars/Knucklessg1)\n\n---\n\n## Contribute\n\nContributions are welcome! Please ensure code quality by executing local checks before submitting pull requests:\n- Format code using `ruff format .`\n- Lint code using `ruff check .`\n- Validate type-safety with `mypy .`\n- Execute test suites using `pytest`\n\n\n\u003c!-- BEGIN agent-os-genesis-deploy (generated; do not edit between markers) --\u003e\n\n## Deploy with `agent-os-genesis`\n\nThis package can be provisioned for you — skill-guided — by the **`agent-os-genesis`**\nuniversal skill (its *single-package deploy mode*): it picks your install method, seeds\nsecrets to OpenBao/Vault (or `.env`), trusts your enterprise CA, registers the MCP\nserver, and verifies it — the same machinery that stands up the whole Agent OS, narrowed\nto just this package. Ask your agent to **\"deploy `vector-mcp` with agent-os-genesis\"**.\n\n| Install mode | Command |\n|------|---------|\n| Bare-metal, prod (PyPI) | `uvx vector-mcp` · or `uv tool install vector-mcp` |\n| Bare-metal, dev (editable) | `uv pip install -e \".[all]\"` · or `pip install -e \".[all]\"` |\n| Container, prod | deploy `knucklessg1/vector-mcp:latest` via docker-compose / swarm / podman / podman-compose / kubernetes |\n| Container, dev (editable) | deploy `docker/compose.dev.yml` (source-mounted at `/src`; edits live on restart) |\n\nSecrets are read-existing + seeded via `vault_sync` — you are only prompted for what's missing.\n\n\u003c!-- END agent-os-genesis-deploy --\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fknuckles-team%2Fvector-mcp","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fknuckles-team%2Fvector-mcp","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fknuckles-team%2Fvector-mcp/lists"}