bigfish
bigfish is the solo open-source developer identity behind rpi (pi-rust), a Rust-native coding-agent runtime and terminal agent. rpi is designed to let developers read and edit code, run commands, manage durable sessions, embed an agent in Rust applications, and extend it through providers and Rust plugins.
At a Glance
- Rust developers and application builders
- Developers who want an embeddable coding-agent runtime rather than only a CLI
- Terminal and automation users seeking a lightweight native coding agent
- Teams building internal developer tools, editor integrations, or headless agent services
- +1 more
AI Tools by bigfish
(1)
rpi
Open Source Rust Coding Agent
Discussions
No discussions yet
Be the first to start a discussion about bigfish
Latest News
rpi v0.9.2 released with a CLI model/footer switching fix and release assets for Linux, macOS, and Windows.
rpi v0.9.1 released with a Windows IME commit-buffering fix.
The repository published a runtime/performance optimization change extracting reusable rpi-runtime and reducing repeated context/message copies.
rpi v0.9.0 released with a fix keeping temporary model switches out of settings.json.
Products & Services
A native Rust CLI/TUI coding agent distributed as a standalone binary, with interactive, one-shot, JSON, RPC/server, and remote-client modes.
A library-first stack including rpi-ai for provider-agnostic messages and streaming, rpi-agent for the agent loop, rpi-tools for read/write/edit/bash tools, rpi-harness for durable sessions and recovery, and rpi-runtime for reusable model/configuration bindings.
A Rust cdylib plugin system with a stable repr(C) ABI for tools, providers, slash commands, event handlers, resources, and child agents; rpi-extensions loads and bridges plugins.
A headless server and thin terminal client allowing the agent, providers, tools, extensions, and session files to run on one machine while a client connects remotely.
Market Position
rpi positions itself as a Rust-native, library-first runtime with a CLI attached, rather than only a terminal application. Its differentiators are embeddability, a first-class Rust plugin ABI, provider-agnostic gateways, durable crash-resumable sessions, offline testability, and a single native binary. The README compares it structurally with Claude Code, Codex CLI, opencode, and aider, while noting that rpi is for users who want to own and compose the agent loop.
Leadership
Founders
bigfish1913 (BigFish)
The GitHub maintainer and Cargo author of the pi-rust/rpi project; the repository credits the package author as BigFish and the GitHub profile shows ownership and ongoing development of pi-rust and related distribution repositories. Public sources identify the developer by the handle rather than a legal name or prior employer.
Founding Story
The developer started rpi while learning Rust and wanting to understand how a complete coding agent works: model requests, streaming output, tool execution, cancellation, and session recovery. The project began as a Rust reimplementation of Mario Zechner's Pi TypeScript SDK, with an initial vision of making the agent reusable as an embeddable Rust library as well as a low-overhead terminal program, distributed as a single native binary.
Business Model
Revenue Model
rpi is MIT-licensed open-source software with free source, binaries, crates, and documentation. The project README also solicits voluntary tips/donations to support continued development and maintenance; no paid subscription or commercial license model is described.
Target Markets
- Rust developers and application builders
- Developers who want an embeddable coding-agent runtime rather than only a CLI
- Terminal and automation users seeking a lightweight native coding agent
- Teams building internal developer tools, editor integrations, or headless agent services
- Extension and plugin authors
- Interactive terminal coding and repository exploration
- Embedding a coding agent inside a Rust application or service
- Automated or headless coding-agent jobs over JSON, RPC, or a remote client
- Offline testing of agent loops and tools with faux providers and in-memory execution environments
- Building Rust plugins that add tools, model providers, commands, events, or resources
- Durable long-running tasks that need session persistence and crash recovery