# AILANG Executor

> AILANG Executor is a free MCP server for developer tools. Write, type-check and run AILANG code in a sandbox, with Z3-proved contracts. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** API, Web
- **Website:** https://ailang.sunholo.com/docs/guides/hosted-executor
- **Repository:** https://github.com/sunholo-data/ailang
- **MCP Registry name:** io.github.sunholo-data/ailang-executor
- **Categories:** developer tools
- **Listing updated:** 2026-11-14

## About

Every call gets a fresh directory with one main.ail file that is deleted afterwards, and the program runs under a fixed policy the caller cannot change. Limits: IO capability only with no network, environment, clock or process access, 60 seconds of run time, 64 KB of source and 30 calls a minute per account.

Sign in uses Google with a verified email; the executor stores your Firebase user ID, the OAuth client and token digests. Logging is one line per call with outcome, duration, code size and a short hash, never the code itself. All tools are read-only except that submit_feedback needs no sign-in.

Besides running code, it includes the AILANG docs toolset: standard library, examples, design docs and benchmarks. It is also presented as the reference example of an AILANG service published as an MCP connector.

## Features

- **Sandboxed runs:** Fresh directory, fixed policy.
- **Contract checking:** Type-check and run Z3 contract proofs.
- **Docs tools:** Stdlib, examples, design docs, benchmarks.
- **Rate limits:** 30 calls per minute per account.
- **Google sign-in:** OAuth through the connector flow.
- **No code logging:** Only hashes and outcomes logged.

## Good for

- Try AILANG without installing
- Let an assistant check a program
- Evaluate one expression
- Report a bug to the AILANG team

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http ailang-executor https://executor.ailang.sunholo.com/mcp/connect/
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://executor.ailang.sunholo.com/mcp/connect/
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "ailang-executor": {
      "url": "https://executor.ailang.sunholo.com/mcp/connect/"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "ailang-executor": {
      "type": "http",
      "url": "https://executor.ailang.sunholo.com/mcp/connect/"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "ailang-executor": {
      "serverUrl": "https://executor.ailang.sunholo.com/mcp/connect/"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "ailang-executor": {
      "type": "http",
      "url": "https://executor.ailang.sunholo.com/mcp/connect/"
    }
  }
}
```

## Tools

- `ailang_builtins`: List AILANG builtin functions with documentation, grouped by module. `search` filters lines case-insensitively; empty returns everything.
- `ailang_check`: Type-check AILANG source. `code` is a whole module; the module line may be omitted or be `module main`. Returns the ai-check JSON report on stdout: check.passed and check.errors (with locations and fixes), and verify: each function's requires/ensures proved with Z3 (verify.results: verified, counterexample with a witness, or inconclusive after 5s per function). Only `pure` functions (or `! {}`) ar
- `ailang_eval`: Evaluate one pure AILANG expression and print it with show(). Builtins are in scope; for stdlib functions (map, filter, ...) use ailang_run with explicit imports.
- `ailang_prompt`: Call this FIRST: the AILANG teaching prompt (syntax rules, templates, stdlib map). AILANG is not in your training data the way Python is; code written without this prompt rarely type-checks.
- `ailang_run`: Run AILANG source in the hosted sandbox. Entry point is `main`, with no arguments. Only the IO capability is granted (println works; FS, Net, Process, Env and AI are refused by name before the program starts), and the run is killed at the policy timeout. stdout is the program's output; stderr carries the policy decision JSON.
- `ailang_versions`: List all AILANG versions the server can answer for.
- `benchmarks_compare`: Two models side by side across every banked run: passes, runs, pass rate, cost and failures by category. Model names are as benchmarks_list gives them.
- `benchmarks_for_model`: Where does a given model pass/fail? Optional version filter. Output: {passes: [...], fails_by_category: {...}, total_cost}
- `benchmarks_list`: List benchmark suites and runs, optionally filtered to a since_version cutoff.
- `changelog_for_version`: What changed in a given release (added/changed/fixed/breaking).
- `design_doc`: Full design doc by slug (e.g. "m-agent-mcp-website"), as markdown. Long documents are cut at 30,000 characters and say so; url opens the whole one.
- `design_docs_list`: List planned/implemented/rejected design docs, optionally filtered by version. state = "planned" | "implemented" | "rejected" | ""; empty = all.
- `docs_nav`: The AILANG docs site's navigation tree for a version, as JSON: section and page titles with their paths.
- `docs_search`: Full-text search of the AILANG documentation for a version (case-insensitive, on page titles and text). Returns the top 10 matches, docs pages before archived teaching prompts, each with its path, title, first 10 headings and an excerpt around the match.
- `effects_catalog`: All declared effects with capability mapping and the version they appeared in.
- `example_for_concept`: Best-fit examples for a free-form concept query ("how do I use effects?"). Filters server-side on why + title; returns up to 5 matches.
- `example_get`: Full source of one example by path (as examples_list gives it, e.g. "runnable/hello.ail"; a leading "examples/" is accepted), at the release.
- `examples_list`: Examples in the release's manifest (examples/manifest.json at its tag). features = comma-separated tags ("effects,records"); empty = no filter. status = "working" | "broken" | "aspirational" | "" ("passing" = working, "failing" = the other two); empty = working only. At most 50 are returned; total says how many matched.
- `executor_whoami`: Which account this connection is signed in as (its user ID) and which app connected it (for example claude.ai or chatgpt.com).
- `install_guide`: Harness-aware install instructions for AILANG. harness: "claude-code" | "gemini-cli" | "codex-cli" | "manual" | "build-from-source" | "" platform: only meaningful for "manual" (e.g. "darwin-arm64", "linux-amd64"); ignored otherwise Empty harness returns ALL harnesses + the remote_mcp config block.
- `limitations_list`: Known limitations (Y-combinator, etc.) optionally filtered by category.
- `onboarding_guide`: Step-by-step guide to getting started with AILANG, for an AI coding agent and for a human developer. role: "agent" | "developer" | "" (both). The guide for both roles is returned whatever role says; pick the part you need.
- `prompt_get`: Fetch the canonical teaching prompt for a version. kind = "" (the full prompt) | "agent" | "devtools".
- `roadmap`: What's shipping next, what's been rejected and why: the roadmap page of the AILANG site, generated from the planned design docs, as markdown.
- `stdlib_module`: Full module: exports + signatures + docstrings.
- `stdlib_modules`: List stdlib modules available in this AILANG version with one-line summaries.
- `stdlib_search`: Find stdlib functions by name, signature, or keyword. Filters the pre-built search index server-side; returns top 20 matches.
- `submit_feedback`: Send a bug report, feature request, documentation gap or limitation about AILANG to the AILANG team. Use it only when the user asks to report something, and leave out personal or confidential information. Categories: bug, feature, docs, limitation. Body up to 10 KB, snippet up to 4 KB. `contact` is optional, for a reply. `package` (vendor/name) sends the report to that AILANG package's maintainers

## FAQ

### What can a program do?

Only IO; no network, environment, clock, process or file access outside the request directory.

### How long can it run?

60 seconds.

### Is my code logged?

No, only a short hash and outcome.

### Does it need sign-in?

Every tool except submit_feedback.

### Is AILANG Executor free?

Yes. AILANG Executor is free to use.

### How do I add AILANG Executor to Claude Code?

Run this command in your terminal: claude mcp add --transport http ailang-executor https://executor.ailang.sunholo.com/mcp/connect/

### What are the best alternatives to AILANG Executor?

The closest alternatives to AILANG Executor are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/ailang-executor
