Configuring LLM access for Pi
Documentation outlines how to sign in with providers, use API keys, connect to local GGUF models, or add custom endpoints to `models.json`, allowing Pi to interact with diverse language models.
All sources
Pi Coding Agent Documentation for AI Agents & MCP
Pi documentation guides AI agents on configuring models, integrating with CLI and MCP servers, and extending its capabilities. It covers developing custom providers, codemodes, and extensions for tailored AI workflows.
- Pages
- 40
- 405
- Tokens
- ~84.8K
- Updated
- Oct 5, 2026
- Monitoring
- Active
Works with
- ChatGPT
- Claude
- Cursor
- MCP clients
Solutions
What AI Agents Can Do With This Source
Use Pi Documentation to complete source-specific tasks, make informed decisions, and follow evidence-backed workflows.
Automating Pi tasks via CLI modes
The source details interactive, print, JSON, and RPC modes, explaining how each handles input and output, allowing agents to select the appropriate interface for scripts and applications.
Developing custom Codemode scripts
Documentation explains writing JavaScript scripts for the `codemode` tool, detailing its sandbox, available tools, and options like `max_output_tokens` and `timeout_ms` to extend model capabilities.
Creating Pi extensions for new functionality
The source provides guidance on developing TypeScript modules that add executable behavior, such as tools, commands, or event handlers, by exporting a default factory that registers capabilities within the Pi process.
Integrating Pi with Model Context Protocol servers
The documentation outlines the quick setup and configuration of MCP servers, detailing how Pi connects to them over stdio or HTTP to make their tools and resources available to the model.
Customizing Pi's terminal keybindings
The Keybindings Reference explains how to assign custom keybindings to named actions by editing `keybindings.json` in the agent directory, allowing users to tailor the terminal UI and application shortcuts.
Coverage
What's Indexed from Pi
A quick look at the main areas Bulkgrid has indexed and keeps ready for AI search, MCP, extraction, and change detection.
Coding Agent
/packages/coding-agent
40 pages
Houses the Pi coding agent package, including its source code, configuration, and documentation, enabling users to interact with models, utilize CLI commands, and extend its capabilities.
Docs
/packages/coding-agent/docs
40 pages
Documents the Pi coding agent's usage, covering CLI commands, integration modes, model selection, configuration, environment variables, and methods for developing extensions.
Advantages
Why Use Bulkgrid for AI-Ready Sources
Bulkgrid turns important websites into fresh, queryable, AI-ready sources without making teams maintain the infrastructure.
Spend Fewer Tokens
Retrieve the right source context instead of sending large pages, pasted docs, or noisy search results into every prompt.
Improve Answer Accuracy
Ground responses in current, source-backed evidence to reduce hallucinations and avoid relying on stale model memory.
Use Fresh Source Data
Keep important websites automatically recrawled and updated so AI is not working from old content.
Query Instantly
Search already-indexed source content immediately instead of crawling pages at question time.
No Infrastructure Maintenance
Bulkgrid handles crawling, rendering, retries, storage, indexing, recrawls, and change detection.
Change Detection
See when pages are added, removed, updated, or fail so teams can react without checking manually.
Give Your Agents Up-to-Date Knowledge from Pi
Search and use knowledge from Pi in ChatGPT, Claude, Cursor, and any MCP-compatible client.