Documentation: docs.toolfront.ai

Source code: https://github.com/statespace-ai/toolfront

ToolFront helps you build and deploy environments for AI agents. Think of environments as interactive directories that agents can explore and take actions in.

environment/
├── index.md
├── pages/
│   ├── text2sql.md
│   ├── document.md
│   └── api.md
├── tools/
│   └─ extract.py
└── data/
    ├── invoices/
    └── logs/

Agents can run commands listed in markdown headers. As they browse files, they will discover these tools and learn how to use them with the --help flag.

Landing Page

---
tools:
  - [date, +%Y-%m-%d]
---

# Landing Page

Add instructions and tools to markdown pages.
- Agents can only run commands in headers
- Links to [pages](./pages) help with navigation

---
tools:
  - [toolfront, database, $POSTGRES_URL]
---

# Text-to-SQL

Build text-to-SQL workflows with the `toolfront database` CLI.
- Agents may run `list-tables`, `inspect-table`, and `query` subcommands
- All queries are restricted to read-only operations

---
tools:
  - [python, tools/extract.py]
---

# Document RAG

Link to [directories](./data) where documents are stored.
- Agents use built-in tools like `read`, `glob`, and `grep` to search files
- Custom tools can be added for data extraction and processing

---
tools:
  - [curl, -X, GET, "https://api.products.com/v1/pricing"]
---

# API Integration

Define API endpoints as executable tools using `curl` commands.
- Agents can call external APIs to fetch live data
- Include environment `$VARIABLES` for authentication

You can launch browsing sessions with ToolFront's Python SDK, or build your own browsing agent with the MCP. Browsing is always powered by your own models.

SDK

from toolfront import Browser

browser = Browser(model="openai:gpt-5")

url = "file:///path/to/environment"

answer = browser.ask("What's our average ticket price?", url=url)
print(answer)

{
  "mcpServers": {
    "toolfront": {
      "command": "uvx",
      "args": ["toolfront", "mcp", "file:///path/to/toolsite"]
    }
  }
}

ToolFront comes with six core tools your agents can use to interact with environments:

• run_command - Execute commands defined in markdown headers
• read - Read the content of a specific file
• tree - View directory structure
• glob - List files matching a glob pattern
• grep - Search files using regex patterns
• search - Find relevant documents using BM25 full-text search*

*search requires indexing environment files.

Installation

To get started, install toolfront using your favorite PyPI package manager.

pip install toolfront

Deploy with ToolFront Cloud

Instantly deploy your environments with ToolFront Cloud.

toolfront deploy ./path/to/environment --api-key "my-api-key"

Would give you a secure environment URL your agents can browse.

answer = browser.ask(..., url="https://cloud.toolfront.ai/user/environment")

Environments deployed with ToolFront Cloud are automatically indexed and get access to the powerful search tool.

Let me search the environment for documents relevant to "ticket pricing API"...

Found 3 relevant pages:
  - ./api/pricing.md (highly relevant)
  - ./guides/analytics.md (relevant)
  - ./examples/queries.md (somewhat relevant)

I'll start by reading ./api/pricing.md...

ToolFront Cloud is currently in open beta. To request access, join our Discord or email esteban[at]kruskal[dot]ai.

Community & Contributing

• Discord: Join our community server for real-time help and discussions
• X: Follow us @toolfront for updates and news
• Issues: Report bugs or request features on GitHub Issues

License

This project is licensed under the terms of the MIT license.