Skip to main content
Glama

VectorSmith

Your vector database, forged into tools an agent can actually use.

Write a tools.yaml. VectorSmith compiles it into typed, tenant-guarded tools — then you either import them in Python or serve them over MCP.

License Python 3.11+ TDS MCP Docs

What it is · How it works · Write YAML · Python · Claude / Codex / Cursor · Try it · Docs


Why this exists

Agents that talk to your invoices, tickets, or catalog usually get one of two bad options:

Typical approach

What goes wrong

Vendor MCP (Qdrant / Pinecone / …)

Cluster admin tools. Upsert, delete, create-collection. The model can wander.

Hand-bind JSON schemas to LangChain / the OpenAI SDK

You re-implement filters, limits, and tenant isolation in Python. Every agent copies it.

“Just embed and search() in the system prompt”

No typed args. No enums. No hidden tenant = acme.

VectorSmith is the third option: the data store stays yours. The tools are a YAML contract. The compiler turns that contract into MCP schemas or in-process tools. The agent never sees the URL, the API key, or the tenant filter.

  you write                         VectorSmith                    the agent sees
─────────────                   ─────────────────                ────────────────
 tools.yaml          ──▶   interpolate → validate → compile  ──▶  search_invoices
 tenant: acme                    Engine stays internal            query, client, status
 ${QDRANT_URL}                                                    (no tenant, no URL)

Related MCP server: openapi-mcp-server

How it works

flowchart LR
  subgraph author["You"]
    Y["tools.yaml"]
    E[".env / ${VAR}"]
  end
  subgraph vs["VectorSmith"]
    L["load + secret lint"]
    V["validate VBxxxx"]
    C["compile schemas + plan"]
  end
  subgraph out["Consume once"]
    P["load_tools() / connect()"]
    M["vectorsmith serve"]
  end
  subgraph hosts["Hosts"]
    A["LangChain · LangGraph · Agents SDK · Anthropic"]
    H["Claude · Codex · Cursor · claude.ai"]
  end
  Y --> L
  E --> L
  L --> V --> C
  C --> P --> A
  C --> M --> H

One file, two doors. Same compiled tools.

Python app

Chat / IDE host

Install

pip install "vectorsmith[qdrant,langchain]"

pip install "vectorsmith[qdrant]" so vectorsmith is on PATH

Call

from vectorsmith import load_tools

vectorsmith serve tools.yaml --name invoices

Process

In-process. No subprocess.

The host spawns the CLI (MCP stdio or HTTP)

Mix-in

Your @tools + Slack/GitHub via an MCP client

Other mcpServers keys sit next to it

You do not import an executor. You do not copy inputSchema into the LLM SDK.


Write a tool, not a prompt

A tool is a name, a description (so the model picks it), a collection, optional text search, parameters the model may pass, and filters it must never see:

tds_version: "1"

connections:
  invoices:
    backend: qdrant
    url: ${QDRANT_URL}              # secrets only here, only as ${VAR}
    api_key: ${QDRANT_API_KEY:-}

tools:
  - name: search_invoices
    kind: search
    description: >
      Search invoices by free text and filter by client, status, or amount.
      Use when the user asks about invoices, billing, or payments.
    target: { connection: invoices, collection: invoices }
    query: { param: query, required: false }
    static_filters:
      - { path: tenant, op: eq, value: acme }    # hidden from the model
    parameters:
      - { name: client, path: client_name, dtype: keyword, op: eq }
      - { name: status, path: status, dtype: keyword, op: in,
          enum: [draft, sent, paid, overdue] }
      - { name: min_amount, path: amount, dtype: float, op: gte }
    output:
      fields: [invoice_id, client_name, status, amount]
      limit_default: 10
      limit_max: 50

vectorsmith init ./demo writes a starter file. The full field list — kinds, operators, pipelines, built-ins, every backend — is in docs/tools-yaml-reference.md.

What the model sees

{
  "name": "search_invoices",
  "description": "Search invoices by free text and filter by client, status, or amount. …",
  "inputSchema": {
    "type": "object",
    "properties": {
      "query": { "type": "string" },
      "client": { "type": "string" },
      "status": {
        "type": "array",
        "items": { "type": "string", "enum": ["draft", "sent", "paid", "overdue"] }
      },
      "min_amount": { "type": "number" },
      "limit": { "type": "integer", "minimum": 1, "maximum": 50, "default": 10 }
    }
  }
}

tenant: acme is not in that schema. The engine ANDs it on every call. Credentials never leave connections.

Kinds you can declare

kind

For

Typical tool

search

Semantic retrieve + filters

search_invoices

lookup

Exact id, limit 1

get_invoice

count

“How many overdue?”

count_invoices

scroll

Filter / page, no ANN

list-style tools

pipeline

Retrieve → post_filter / group_by / sort / project

top-N per client

Built-ins (search_<connection>, get_<connection>_by_id, …) are opt-in on the connection. Turn them off if you already named a user tool the same way.


In your agent (Python)

pip install "vectorsmith[qdrant,langchain]"
from vectorsmith import load_tools
from langchain.agents import create_agent

tools = load_tools("tools.invoices.yaml", "tools.tickets.yaml")
agent = create_agent("openai:gpt-4.1", tools)
# … await tools.aclose()

Same YAML, other stacks:

from vectorsmith.langgraph import load_tools      # create_react_agent / ToolNode
from vectorsmith.openai_agents import load_tools  # Agent + Runner
from vectorsmith.anthropic import load_tools      # messages.create(tools=vs.tools)
from vectorsmith import connect                   # await vs.call("search_invoices", {…})

Extra

Import

vectorsmith[langchain]

from vectorsmith import load_tools

vectorsmith[langgraph]

same tools; LangGraph graph

vectorsmith[openai-agents]

from vectorsmith.openai_agents import load_tools

vectorsmith[anthropic]

from vectorsmith.anthropic import load_tools

Worked apps: examples/langchain_agent · langgraph_agent · openai_agents · anthropic_agent.


In Claude, Codex, Cursor

Those products cannot import vectorsmith. They spawn a process. Point them at serve with the same YAML.

{
  "mcpServers": {
    "invoices": {
      "command": "vectorsmith",
      "args": ["serve", "tools.invoices.yaml", "--name", "invoices"]
    }
  }
}

Codex is TOML (~/.codex/config.toml), not JSON. Claude Code uses .mcp.json — it does not read the Desktop file.

Host

Config

Guide

Claude Desktop

claude_desktop_config.json

docs/integrations/claude-desktop.md

Claude Code

.mcp.json / claude mcp add

docs/integrations/claude-code.md

OpenAI Codex

~/.codex/config.toml

docs/integrations/openai-codex.md

Cursor

.cursor/mcp.json

docs/integrations/cursor.md

claude.ai

serve --http --auth builtin

docs/quickstart-selfhost.md

Copy-paste snippets: examples/mcp_hosts/. Slack, GitHub, filesystem stay separate servers — coexistence.


Stores

backend on a connection is one of six shipped adapters. Full matrix (extras, hybrid, nested paths): vector stores.

qdrant · pgvector · chroma · pinecone · weaviate · milvus

pgvector can run in table mode (no vector column) for lookup / count / scroll. Hybrid search is capability-gated (Qdrant / Weaviate / Milvus / Pinecone) and checked with validate --live.


Try it

The invoice example is a tools.yaml plus an env file. Copy .env.example and set QDRANT_URL to your cluster before validate / test / serve.

# clone, then:
uv sync

uv run vectorsmith validate examples/qdrant_invoices/tools.invoices.yaml \
  --env-file examples/qdrant_invoices/.env.example

uv run vectorsmith test examples/qdrant_invoices/tools.invoices.yaml search_invoices \
  --args '{"query":"Globex invoice","limit":3}' \
  --env-file examples/qdrant_invoices/.env.example

uv run vectorsmith serve examples/qdrant_invoices/tools.invoices.yaml --name invoices \
  --env-file examples/qdrant_invoices/.env.example

Tickets are a second file / second MCP name: tools.tickets.yaml--name tickets.

Example walkthrough


CLI

Command

Does

init

Write a starter tools.yaml + .env.example

validate

Compile + lint. --live pings the store. --strict fails on warnings

test

Call one compiled tool without serving

serve

MCP stdio (Desktop / Codex / Cursor; --watch on by default) or --http HOST:PORT (no watch). Default HTTP --auth is builtin (needs https --public-url). Localhost HTTP: --auth none.

introspect

Collection / field metadata to --out (default schema.json). Requires --connection.

drafts / approve

drafts list|reject NAME. approve NAME [--file tools.yaml] promotes into that file. Drafts live in ./tools.drafts.yaml (process cwd).

auth

rotate-secret | revoke for builtin HTTP OAuth

validate exits 0 / 1 (--strict warnings) / 2 (errors). test and introspect use 3 on a live failure. serve --http --auth none off localhost exits 3.


Documentation

kjgpta.github.io/vectorsmith is the rendered manual (Material for MkDocs). Source is docs/.

I want to…

Go here

Get a tool working in five minutes

Getting started

See which vector stores ship

Vector stores

Understand every tools.yaml field

YAML reference

Plug into Claude, Codex, Cursor, LangChain, …

Integrations

Look up a CLI flag

CLI

Call tools from Python

Python API

Fix Desktop disconnect / env / HTTP auth

FAQ

Copy a host config

examples/mcp_hosts

See agent apps

examples/


Develop

uv sync
uv run ruff check .
uv run pytest -m "not conformance"
uv run lint-imports

Workspace: packages/core (vectorsmith_core, unpublished) · packages/cli (published vectorsmith). Core must not import the CLI.

Contributing · Support · Security · Changelog · Code of conduct


Apache-2.0 · LICENSE · NOTICE

Forge the tools. Keep the store.

Install Server
A
license - permissive license
A
quality
A
maintenance

Maintenance

Maintainers
Response time
Release cycle
1Releases (12mo)
Commit activity

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Servers

  • A
    license
    A
    quality
    D
    maintenance
    Enables AI-powered generation of production-ready CTP (ConveniencePro Tool Protocol) tools from natural language descriptions, including tool definitions, implementations, tests, and TypeScript validation.
    5
    12
    MIT
  • A
    license
    -
    quality
    C
    maintenance
    Converts any OpenAPI/Swagger API specification into MCP tools that AI assistants can use to interact with the API.
    24
    7
    MIT
  • A
    license
    B
    quality
    C
    maintenance
    Transforms OpenAPI definitions into MCP tools for seamless LLM-API integration.
    8
    39
    1
    MIT

View all related MCP servers

Related MCP Connectors

  • Point Gecko at an OpenAPI spec; get first-call-correct, auth-hidden agent tools.

  • Reliable async execution for agent tool calls: schema gating, retries, idempotency, audit trail.

  • 33 tools that make AI write, implement, and verify intent against explicit, testable constraints.

View all MCP Connectors

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/kjgpta/vectorsmith'

If you have feedback or need assistance with the MCP directory API, please join our Discord server