Tools · agents
How to use AI Code Sandbox from your AI agent
● ready to runagents
An agents tool your agent can run. It takes inputs such as Agent skills, Node.js dependencies and Python requirements, and returns its results to your agent. One prepaid apifare balance pays for this and every other capability your agent uses — under per-agent caps, a kill switch, and a signed receipt for every call.
What you need
An apifare key with credits. That's it: get a key (new accounts start with 500 free credits), connect your agent with its setup guide, and ask it to run AI Code Sandbox.
Inputs it takes
| field | what it sets | example |
|---|---|---|
agentSkills | Agent skills | |
nodeDependencies | Node.js dependencies | |
pythonRequirements | Python requirements | |
mcpConnectors | MCP connectors | |
initBashScript | Setup script | |
envVars | Environment variables |
Your agent fills these in from what you ask; you rarely need to name them.
Price
Reserve 5 credits per run, settled to the run's true cost (ceiling 20 credits; 1 credit = $0.01). Failed runs are refunded. The exact billed number always comes from /capabilities.
Run it
Once your agent is connected, ask it:
Use apifare to run "AI Code Sandbox" for me.Your agent finds it with search_apis and runs it with proxy_call (in apifare-mcp). Over HTTP: GET /v1/search?q=… with Authorization: Bearer <token>, then POST the result's proxy_route with the input JSON. Full reference: the docs · for agents: this page as Markdown.
Give this to your agent
set up https://apifare.com/SKILL.md — and it takes it from there: key, config, first call. $5 of calls free.
No agent open right now? ① Get a key — $5 of calls free, then ② give it to your agent (setup guides).
Tool data © its developer. apifare routes and meters the call; you are responsible for complying with the platform's terms (terms).