# Compute Connector [Health: Active]

**Category:** 👨‍💻 Code Execution  
**Repository:** https://compute.cloudforbots.com/  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/compute-connector

## Description
Low on memory? Your bot sends heavy jobs to cloud sandboxes on your own provider keys. Flat fee.

## Tools
Capabilities this server exposes over MCP:

- **check_memory** — Before a heavy job, send how much memory the machine your agent runs on has free (free -m, os.freemem()) and what you are about to run. Get a memory warning and a verdict: run_local, or offload with the provider to use. Free watch-only logs would_offload for the weekly report; nothing runs.
- **run_job** — Pro: run a shell command in a fresh cloud sandbox on your own provider account (vercel (Vercel Sandbox, default), e2b, daytona, modal, workers (Workers Sandbox)) and get the exit code and the last 16 KB of stdout and stderr back. Send small input files with it. The sandbox is created for the job and stopped right after, so nothing idles on your bill. Free watch-only logs would_offload and runs nothing.
- **save_provider_key** — Pro, owner only: save one provider key, sealed with AES-256-GCM before it reaches the database. Providers: vercel (Vercel Sandbox, default), e2b, daytona, modal, workers (Workers Sandbox). Vercel needs team_id and project_id; Modal and Workers Sandbox need bridge_url (see provider_setup). The key is checked with a read-only call first and is never returned or logged. Free stores nothing.
- **list_providers** — Every provider this connects to, how each one runs, and (with your key) the provider keys you saved: name, provider, fingerprint, default, last used. Never shows a key.
- **delete_provider_key** — Pro, owner only: delete a saved provider key by name. Jobs on that key stop working at once.
- **provider_setup** — Step-by-step setup for one provider: where to get the key, what to pass to save_provider_key, links to its docs and price page, and for Modal or Workers Sandbox the bridge code to deploy on your own account.
- **weekly_report** — Plain-English summary of the last 7 days: memory checks, low-memory warnings, jobs that would have been offloaded (free) or ran (Pro) by provider, minutes and failures. Read-only; sends nothing. Needs an API key.
- **usage_and_upgrade** — Show today's usage for your key or IP and the $9/month upgrade link. Free tier: 500 calls/day. Not metered.

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What Compute Connector MCP server does

Compute Connector helps an agent decide whether a task should run on its local machine or in a cloud sandbox. The `check_memory` tool accepts available-memory information and a description of the planned work, then returns a warning and either a `run_local` or offload recommendation.

For Pro usage, `run_job` starts a fresh sandbox with a selected provider, executes a shell command, optionally uploads small input files, and returns the exit code along with the last 16 KB of standard output and standard error. The sandbox is stopped after the job, so the job does not leave an idle runtime running on the provider account.

Supported providers are Vercel Sandbox, e2b, Daytona, Modal, and Workers Sandbox. Vercel is the default provider.

## How it works

The server uses provider credentials saved for the account owner. `save_provider_key` validates a credential with a read-only provider call, encrypts it with AES-256-GCM before it reaches the database, and does not return or log the key. Saved credentials are referenced by name rather than exposed to the agent. `list_providers` shows provider names, fingerprints, default status, and last-use information without displaying secrets.

Free operation is watch-only for execution-related workflows: memory checks can record jobs that would have been offloaded, and `run_job` does not execute a command. Pro operation runs the requested job. Weekly reporting summarizes memory checks, warnings, would-offload events or completed jobs, provider usage, minutes, and failures for the previous seven days.

## Setup and configuration

Provider credentials are configured through the MCP tools rather than by returning secrets to the caller. Use `provider_setup` to get provider-specific instructions, credential locations, save parameters, documentation links, and pricing links.

Vercel requires both `team_id` and `project_id` when its key is saved. Modal and Workers Sandbox require a `bridge_url`; their bridge code is provided for deployment on the user’s own account. Provider keys can be removed by name with `delete_provider_key`, after which jobs using that key stop working.

## Tools and capabilities

- `check_memory`: evaluates available memory and the planned task.
- `run_job`: executes a shell command in a temporary provider sandbox for Pro users.
- `save_provider_key`: validates and stores one encrypted provider credential for the owner.
- `list_providers`: lists providers and non-secret saved-key metadata.
- `delete_provider_key`: removes a saved provider key.
- `provider_setup`: provides setup details for a selected provider.
- `weekly_report`: returns a read-only seven-day activity summary.
- `usage_and_upgrade`: shows daily usage and the upgrade link.

## Limitations and notes

The free tier allows 500 calls per day and is not metered. Free execution workflows only record what would have happened; they do not run cloud jobs. `weekly_report` requires an API key, while `usage_and_upgrade` can show usage for an API key or IP. Removing a provider key immediately prevents jobs that depend on it from working. Job output is limited to the final 16 KB of stdout and stderr.

_Full upstream README: https://allmcps.com/mcp/compute-connector/readme_