Use the baseten skill with the Baseten backend and documentation Model Context Protocol (MCP) servers to work with Baseten from your coding agent.
You can ask your agent to:
- Diagnose deployment failures by inspecting logs and proposing fixes.
- Promote deployments, update autoscaling, and test inference.
- Check the status of models in your workspace.
- Find product guidance in the Baseten documentation.
- Write deployment configurations and API clients.
See evaluation results for measured performance and coverage.
- To interact with your Baseten workspace, create an API key with management permissions. Use a dedicated key so you can revoke it without affecting other work.
- Install Node.js 18 or later to run the installation tools.
Set the BASETEN_MCP_KEY environment variable in your shell. Replace ... with your API key.
On Linux or macOS:
export BASETEN_MCP_KEY=...On Windows with PowerShell:
$env:BASETEN_MCP_KEY = "..."On Windows with Command Prompt:
set BASETEN_MCP_KEY=...Then paste the following instructions into your agent:
Install the Baseten agent toolkit following instructions from `github.com/basetenlabs/baseten-skills`.
- use `npx skills add` and `npx add-mcp`
- all global and for all agents (`-g -y`)
- The baseten backend server needs auth header `Authorization: Bearer $BASETEN_MCP_KEY` (or `Authorization: Bearer $env:BASETEN_MCP_KEY` in PowerShell). Run the install commands in a shell that has `BASETEN_MCP_KEY` set so the env var expands; don't read or inline the key value.
On Linux or macOS:
export BASETEN_MCP_KEY=...
npx skills add basetenlabs/baseten-skills -g -y
npx add-mcp https://api.baseten.co/mcp -g -y --header "Authorization: Bearer ${BASETEN_MCP_KEY}"
npx add-mcp https://docs.baseten.co/mcp -n "baseten_docs" -g -yOn Windows with PowerShell:
$env:BASETEN_MCP_KEY = "..."
npx skills add basetenlabs/baseten-skills -g -y
npx add-mcp https://api.baseten.co/mcp -g -y --header "Authorization: Bearer $env:BASETEN_MCP_KEY"
npx add-mcp https://docs.baseten.co/mcp -n "baseten_docs" -g -y-ginstalls globally.-yconfirms installation for all detected agents.- If your agent supports environment variable interpolation, reference
$BASETEN_MCP_KEYin its MCP configuration. - Some agents prompt before reading skill reference files (they live outside your workspace). In Claude Code,
allow these reads in
settings.json:"permissions": { "allow": ["Read(~/.claude/skills/**)"] }.
To create deployments, install the Baseten CLI. baseten model push is the default path for creating and updating deployments:
brew tap basetenlabs/baseten
brew install basetenInstall the Truss CLI as well only when you author Chains, or when you want the truss commands directly. You can install the skill, either MCP server, and the CLIs separately.
After installation, follow your agent's instructions to reload its configuration. If your agent supports /mcp or /mcps, use that command to check the MCP connections. If the backend connection fails, verify BASETEN_MCP_KEY in the agent's configuration.
Ask your agent to explain Baseten documentation, create a deployment, or manage your workspace. Agents that support automatic skill selection can load the skill when needed. If your agent supports skill commands, you can also invoke /baseten.
The September 9 evaluation compared the refreshed skill, the June skill, and no skill on the same model with both MCP servers enabled.
| Configuration | Mean assertion pass rate across 21 tasks |
|---|---|
| No skill | 85.7% |
| June skill | 85.7% |
| Initial September skill | 81.7% |
This 63-execution sweep used one repetition per configuration and task. It did not establish a quality improvement. After correcting client guidance and assertions, all six follow-up executions on tasks 50 and 51 passed. The final skill revision has only that targeted coverage; it has not received a full 21-task sweep. Rubric errors and shared fixture history limit the initial comparison.
The evaluation guide covers running a new comparison and indexes the dated reports, including earlier experiments with different models and suites.