Versions:

0.85.1
0.85.0
0.84.4
0.84.3
0.84.2
0.84.1
0.84.0
0.83.0
0.82.1
0.82.0
0.81.1
0.81.0
0.80.10
0.80.9
0.80.8
0.80.7
0.80.6
0.80.5
0.80.3
0.80.2
0.80.1
0.80.0
0.79.10
0.79.9
0.79.8
0.79.7
0.79.6
0.79.5
0.79.4
0.79.3
0.79.2
0.79.1
0.79.0
0.78.1
0.78.0
0.77.0
0.76.0
0.75.5
0.75.4
0.75.3
0.75.1
0.75.0
0.74.1
0.74.0
0.73.1
0.73.0
0.72.1
0.72.0
0.71.1
0.71.0
0.70.6
0.70.5
0.70.3
0.70.2
0.70.1
0.70.0
0.69.0
0.68.1
0.68.0
0.67.68
0.67.67
0.67.6
0.67.5
0.67.4
0.67.3
0.67.2
0.67.1
0.66.1
0.66.0
0.65.2
0.65.1
0.65.0
0.64.0
0.63.2
0.63.1

Pi is an AI coding assistant developed by EARENDIL INC., currently at version 0.85.1 with 75 versions recorded in its release history. Designed as a minimal terminal coding harness, pi equips the AI with read, bash, edit, and write tools, positioning it within the developer tools and AI-assisted programming category. Its core purpose is to adapt to the user's workflows rather than forcing users to adapt to it, and it achieves this without requiring developers to fork or modify pi internals. Customization is handled through TypeScript Extensions, Skills, Prompt Templates, and Themes, all of which can be bundled into Pi Packages and shared with others via npm or git, making community distribution a first-class part of the ecosystem. Pi ships with powerful defaults while deliberately skipping built-in features such as sub agents and plan mode; instead, users can ask pi to build the functionality they want or install a third-party pi package that matches their preferred workflow. The software runs in four distinct modes to cover a range of use cases: an interactive mode for direct terminal work, print or JSON output modes for scripting and pipeline integration, an RPC mode for process integration, and an SDK for embedding pi within users' own applications. This versatility makes pi suitable for interactive coding sessions, automated tooling, and deeper application integration alike. A real-world example of SDK integration can be found in the openclaw/openclaw project, which demonstrates how pi can be embedded into a larger system. Overall, pi targets developers seeking a lightweight, extensible AI coding harness that prioritizes composability and user-defined workflows over a monolithic feature set.

Tags:

agent 145

agentic 74

ai 414

chatbot 104

code 122

coding 120

large-language-model 144

llm 217

programming 146