Skip to main content

lambda-cloud-mcp

Lets an AI agent check GPU prices and availability on Lambda Cloud, find the cheapest free GPU, and see your instances.

Who it's for: people who rent GPUs on Lambda and use an AI agent such as Claude Code or Claude Desktop. What you get: the agent answers "is an H100 free, where, and how much per hour?" in one step. It is read-only by default. Launching costs money, so it is off until you switch it on, and it has a price limit.

Quick start

  1. Get an API key from the Lambda Cloud API keys page: https://cloud.lambda.ai/api-keys
  2. Install uv (or use pip, see below).
  3. Add the server to Claude Code, putting your key in place of your-key: claude mcp add lambda-cloud -e LAMBDA_API_KEY=your-key -- uvx --from git+https://github.com/tonydzi/lambda-cloud-mcp lambda-cloud-mcp

Done when: you ask your agent "what is the cheapest available GPU on Lambda?" and it answers with a price per hour.

Not affiliated with Lambda. Tested against mocked API responses; a live test is still pending.

Want an AI agent to do it for you? Paste this to Claude Code / Codex / Cursor: "Read the 'For AI agents' section of https://github.com/tonydzi/lambda-cloud-mcp and set it up for me."

For AI agents

Preconditions

  • Python 3.10 or newer, and uv (for uvx) or pip.
  • A Lambda API key. Ask the human for it. Never invent one.
  • Ask the human whether launch/restart/terminate (paid, partly irreversible) should be enabled. Default answer: no.

Install

Either, with uv (no install step):

uvx --from git+https://github.com/tonydzi/lambda-cloud-mcp lambda-cloud-mcp

or with pip:

pip install git+https://github.com/tonydzi/lambda-cloud-mcp

The console command is lambda-cloud-mcp.

Configure

  • LAMBDA_API_KEY (required, from the human): read only from this env var.
  • LAMBDA_MCP_ALLOW_WRITE=1 (optional, only with the human's explicit yes): enables launch_instance, restart_instances, terminate_instances.
  • Register with Claude Code (read-only):
    claude mcp add lambda-cloud -e LAMBDA_API_KEY=<key-from-human> -- uvx --from git+https://github.com/tonydzi/lambda-cloud-mcp lambda-cloud-mcp
    
  • Claude Desktop: add an mcpServers.lambda-cloud entry with command: uvx, args: ["--from", "git+https://github.com/tonydzi/lambda-cloud-mcp", "lambda-cloud-mcp"] and env.LAMBDA_API_KEY. Full JSON is in Details.

Verify

Step 1, no key needed, from a clone (mocked HTTP, no real calls). Works on Windows, macOS and Linux:

uv run --extra test pytest -q

Expected: 14 passed, exit code 0 (real run 2026-11-13). Without uv: pip install -e ".[test]" then python -m pytest -q. This proves the code logic against mocked answers only. It does not prove your key works or that the live API still matches.

Step 2, needs the key and the server registered (see Configure): restart the agent, then ask it to call list_instance_types. A working setup returns a list of objects with name, gpu_description, price_usd_per_hour and regions_with_capacity, cheapest first. A wrong key returns an error text like Lambda API 401 on /instance-types: global/invalid-api-key. The live call has not been run by the author yet, so treat a mismatch as news and report it.

Uninstall / rollback

claude mcp remove lambda-cloud, then pip uninstall lambda-cloud-mcp if you used pip. Remove the key from your shell or config. If write mode was on, remove LAMBDA_MCP_ALLOW_WRITE.

Rules for the agent

  • Never print, log or commit the API key.
  • Do not set LAMBDA_MCP_ALLOW_WRITE=1 unless the human said yes in this conversation.
  • Never launch or terminate instances without the human naming the exact instance and price limit. Termination is irreversible.
  • Lambda rate-limits to about 1 request per second (launch: 1 per 12 seconds). Do not loop calls.

Details

An MCP server for the Lambda Cloud API. It lets an AI agent (Claude Code, Claude Desktop, or any MCP client) check GPU prices and capacity, find the cheapest GPU that is available right now, and see your instances.

If you turn it on, the agent can also launch, restart and terminate instances, with a hard price limit.

Not affiliated with Lambda. Tested against mocked responses built from the public API docs; live test pending.

Why

Lambda has a clean REST API but, as of September 2026, no official MCP server. Agents that train or serve models need to answer simple questions fast: "Is there an H100 free? Where? How much per hour?" This server answers them in one tool call.

Tools

Read-only (always on):

Tool What it does API call
list_instance_types Types with price in USD/hour, specs, regions with capacity. Sorted cheapest first. only_available=true hides sold-out types. GET /api/v1/instance-types
cheapest_available_instance Cheapest type with capacity now. Filters: min_gpus, gpu_contains (e.g. "H100"), region, max_hourly_usd. GET /api/v1/instance-types
list_instances Your running instances GET /api/v1/instances
get_instance One instance by id GET /api/v1/instances/{id}
list_ssh_keys SSH keys in your account GET /api/v1/ssh-keys
list_file_systems Persistent file systems GET /api/v1/file-systems
list_images Available machine images GET /api/v1/images
list_regions Regions GET /api/v1/regions

Write tools (cost money, only with LAMBDA_MCP_ALLOW_WRITE=1):

Tool What it does API call
launch_instance Launch one instance. Requires max_hourly_usd. Refuses if the price is higher, or if the region has no capacity. POST /api/v1/instance-operations/launch
restart_instances Restart instances by id POST /api/v1/instance-operations/restart
terminate_instances Terminate instances by id. Irreversible. POST /api/v1/instance-operations/terminate

Install

You need an API key from the Lambda Cloud API keys page.

Run with uv (no install step):

LAMBDA_API_KEY=your-key uvx --from git+https://github.com/tonydzi/lambda-cloud-mcp lambda-cloud-mcp

Or with pip:

pip install git+https://github.com/tonydzi/lambda-cloud-mcp
LAMBDA_API_KEY=your-key lambda-cloud-mcp

Python 3.10 or newer.

Configure

Claude Code

Read-only:

claude mcp add lambda-cloud -e LAMBDA_API_KEY=your-key -- \
  uvx --from git+https://github.com/tonydzi/lambda-cloud-mcp lambda-cloud-mcp

With launch / terminate enabled:

claude mcp add lambda-cloud -e LAMBDA_API_KEY=your-key -e LAMBDA_MCP_ALLOW_WRITE=1 -- \
  uvx --from git+https://github.com/tonydzi/lambda-cloud-mcp lambda-cloud-mcp

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "lambda-cloud": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/tonydzi/lambda-cloud-mcp", "lambda-cloud-mcp"],
      "env": {
        "LAMBDA_API_KEY": "your-key"
      }
    }
  }
}

Add "LAMBDA_MCP_ALLOW_WRITE": "1" to env only if you want the agent to spend money.

Safety

  • Read-only by default. Without LAMBDA_MCP_ALLOW_WRITE=1 the write tools are not even shown to the agent. They also check the flag again when called. Only the exact value 1 turns it on.
  • Price limit on launch. launch_instance has no default for max_hourly_usd. The server reads the current price from the API and refuses if it is higher. Price is per instance per hour, as returned by the API (price_cents_per_hour).
  • Capacity check. Launch is refused if the region is not in regions_with_capacity_available.
  • One instance per launch call. No bulk launches.
  • Key handling. The key is read only from LAMBDA_API_KEY, sent as Authorization: Bearer, and never logged or returned in errors.
  • API errors come back as readable text with Lambda's error code, e.g. Lambda API 401 on /instances: global/invalid-api-key.

Lambda rate-limits the API to about 1 request per second, and launch to 1 request per 12 seconds.

Source of truth

Every endpoint and field comes from Lambda's public OpenAPI spec, version 1.10.0:

Development

uv run --extra test pytest -q

Tests use respx to mock HTTP. No real API calls, no key needed.

Environment variables

Variable Required Meaning
LAMBDA_API_KEY yes Your Lambda Cloud API key
LAMBDA_MCP_ALLOW_WRITE no 1 enables launch / restart / terminate
LAMBDA_API_BASE_URL no Override base URL (default https://cloud.lambda.ai)

License

MIT

— Anton Dziatkovskii · github.com/tonydzi

READ THIS WITH AI

One click and an agent reads the repo, pulls out the patterns and helps you apply them to your own work.

Codex - open ChatGPT - open Claude - open

Copy the prompt (works in any agent: Gemini, Grok, a local model, your own CLI)
Read this repo: https://github.com/tonydzi/lambda-cloud-mcp (“lambda-cloud-mcp” - MCP server for the Lambda Cloud API: GPU prices, live capacity, cheapest-available lookup, and env-gated launch/terminate with a price cap. Read-only by default). Work out what problem it actually solves, pull out the reusable patterns and help me apply them to my own setup. Start by asking what I am working on.

— TonyDzi, Palo Alto AI Research Lab · second brain, agent coordination, persistent memory: github.com/tonydzi


🧩 One piece of a working system

This repository is one piece lifted out of a live operation: one engineer running a fleet of machines, built with Claude as implementation collaborator; the machines reach consensus with each other and wake the human only for money or the irreversible. It was extracted after it survived production, not written as a demo — and it runs on its own: nothing here phones home to the rest.

See how the whole thing fits together → SYSTEM.md

AI contributors

This project is built by a human + AI team, and the git log says so: Claude writes most of the code, Codex and Grok review it, Gemini feeds the research. Each is credited on a commit only if its output changed that commit's content — no decorative credits. Lab-wide policy, one source for every repo: AI-CONTRIBUTORS.md.

Metadata

Release files for lambda-cloud-mcp 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for lambda-cloud-mcp 0.1.0
File Size Uploaded Provenance Actions
lambda_cloud_mcp-0.1.0.tar.gz 15.3 kB With provenance Details for lambda_cloud_mcp-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for lambda-cloud-mcp 0.1.0
File Interpreter ABI Platform Actions
lambda_cloud_mcp-0.1.0-py3-none-any.whl Python 3 none any Details for lambda_cloud_mcp-0.1.0-py3-none-any.whl

Total release size: 26.6 kB

Release files / lambda_cloud_mcp-0.1.0.tar.gz

Download URL lambda_cloud_mcp-0.1.0.tar.gz
Size 15.3 kB
Tags Source
SHA-256 checksum
How to use checksums
db7930a206390c67485888bcecfbaa2aa58115ff7399bad4fe6692db3c02b09d
BLAKE2b-256 checksum
How to use checksums
f9b908a54311adfcbdcd4b04771226fa9dc49a866c41d17e5ae915097fd667bd
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Release files / lambda_cloud_mcp-0.1.0-py3-none-any.whl

Download URL lambda_cloud_mcp-0.1.0-py3-none-any.whl
Size 11.4 kB
Tags Python 3
SHA-256 checksum
How to use checksums
fe5e6b00ef13bf53ba3bde2e2e2ecb4fe11559edeba14c9fb9260ae1019e4a7b
BLAKE2b-256 checksum
How to use checksums
0641b5bda79daab13cd4fa45fe688ddac66a2fe7a4ebbad6ac69dddf75892ab8
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Security for lambda-cloud-mcp 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for lambda_cloud_mcp-0.1.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files Full provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page