query_example_code_snippet_index_toolLABSCO SUMMARY
Feature code written against the real SDK, not a recollection.
The two indexes do most of the work — find a usage pattern, then confirm the signature — so the assistant writes Tecton code the way the docs say rather than the way it remembers. The metrics tool only appears once an API key is configured.
What it is
A retrieval layer over Tecton's own material for coding assistants: an example-code index, a documentation index, the SDK reference whole or by symbol, and a live metrics query.
What you get
- Tecton code examples found by vector search, for the pattern you are about to write
- Documentation snippets retrieved for a question, straight from the official docs
- The complete Tecton SDK reference, all classes and functions
- The reference for a named list of classes or functions
- Point-in-time system metrics, human-readable or as raw OpenMetrics
Requirements
A Tecton feature repository to work in. Setting an API key in the server's environment registers the API-backed tools, including the metrics query.
Setup effort
Build from source — git clone https://github.com/tecton-ai/tecton-mcp.git
USE CASES
01
Writing a feature view without guessing the API
Find an existing example of the pattern, then check the exact signature before the code is written.
02
Checking what a class or function actually takes
Targeted lookups for named symbols, or the full reference when the shape of the SDK is the question.
03
Answering a Tecton question from the docs
Documentation snippets retrieved for the question at hand, rather than a link to go and read.
04
Seeing how the platform is behaving right now
System metrics come back point-in-time, so a performance question is answered from the deployment rather than from assumption.
5 tools
Listed in project docs. Your client may show a different list.
query_documentation_index_toolget_full_tecton_sdk_reference_toolquery_tecton_sdk_reference_toolquery_tecton_metrics_toolBEFORE YOU INSTALL
What to know first
Tools you get
5 tools named in the project's documentation. Setting TECTON_API_KEY registers more tools, named after the Feature Services in your own Tecton deployment, plus the Metrics API tool; query_tecton_metrics_tool may not work until the key is set.
INSTALLATION
Connect your client
You clone this one yourself.
Server setup
TERMINAL
brew install uv
git clone https://github.com/tecton-ai/tecton-mcp.git
cd tecton-mcp
MCP_SMOKE_TEST=1 uv --directory <YOUR_PATH_TO_LOCAL_CLONE> run mcp run src/tecton_mcp/mcp_server/server.pyLog into your Tecton cluster to authenticate the Tecton SDK used by the Tecton MCP server:
TERMINAL
tecton login <YOUR_CLUSTER>.tecton.aiConfigure Cursor (or any other MCP client) with the MCP server — each client’s configuration is under MCP Clients
Optional, for your feature repository: symlink Tecton’s Cursor rules into it (for Claude Code, symlink CLAUDE.md the same way):
TERMINAL
ln -s <YOUR_PATH_TO_LOCAL_CLONE>/.cursor <YOUR_PATH_TO_TECTON_FEATURE_REPO>/.cursorMCP Clients
Cursor, Augment Code
JSON
{
"mcpServers": {
"tecton": {
"command": "uv",
"args": [
"--directory",
"<YOUR_PATH_TO_LOCAL_CLONE>",
"run",
"mcp",
"run",
"src/tecton_mcp/mcp_server/server.py"
]
}
}
}Claude Code
CLAUDE CODE · TERMINAL
claude mcp add-json tecton-mcp '{
"command": "uv",
"args": [
"--directory",
"<YOUR_PATH_TO_LOCAL_CLONE>",
"run",
"mcp",
"run",
"src/tecton_mcp/mcp_server/server.py"
]
}'
