Earendil Inc.
Earendil crafts software and open protocols to strengthen human agency, bridge division and ignorance, and cultivate lasting joy and understanding. Its current work centers on adaptable AI agent software, including Pi and Lefos.
At a Glance
- Individual software developers
- Open-source developers and agent-framework builders
- Engineering teams and enterprises
- Small businesses and companies seeking customizable AI systems
- +1 more
AI Tools by Earendil Inc.
(1)
Pi Coding Agent
Extensible Terminal AI Coding Agent
Discussions
No discussions yet
Be the first to start a discussion about Earendil Inc.
Latest News
Earendil published What is a Harness?, explaining agent harnesses and positioning Pi as a neutral, open and user-owned harness.
Earendil Engineering published How Compaction Works in Pi, detailing automatic/manual context compaction and preservation of recent work.
Earendil published Pi, Minimal and Performant, highlighting Databricks and Shopify evidence for Pi's cost, extensibility and performance advantages.
Earendil Engineering published The Session You Cannot Take With You, arguing for portable, inspectable and user-owned AI session transcripts.
Products & Services
An MIT-licensed, open-source AI agent toolkit and coding harness with a unified LLM API, agent loop, terminal UI, coding-agent CLI and SDK/RPC/print modes. It is deliberately minimal, model-neutral, extensible and self-editable, with user-owned local sessions.
A machine entity/AI assistant built on the Pi harness, designed to be capable and trustworthy. It communicates primarily over email (and, according to the company profile, Telegram), supports collaboration with co-creators, and can translate messages, edit drafts, analyze documents and schedule recurring tasks; Google Workspace access can create documents, retrieve information and schedule calendar invitations.
An Earendil Works open-source Pi extension/project for transcription.
Market Position
Earendil positions Pi against closed or more heavily orchestrated coding agents such as Claude Code and Codex by emphasizing minimalism, lower context overhead, model/provider neutrality, extensibility, local ownership and open licensing. It positions Lefos as a more deliberate, personal and trustworthy machine entity for communication rather than an engagement- or throughput-maximizing assistant. Its open-core strategy differentiates it from closed AI-lab applications while leaving room for commercial enterprise/cloud features.
Leadership
Founders
Armin Ronacher
Austrian open-source software creator and longtime engineering/team builder; creator of Flask and Jinja and, before Earendil, spent about a decade at Sentry.
Colin Daymond Hanna
Entrepreneur and product-oriented technology leader from San Francisco; previously worked in finance, including as Principal at Balderton Capital.
Executive Team
Armin Ronacher
Co-founder
Creator of Flask and Jinja; open-source software creator and former Sentry engineering/team leader.
Colin Daymond Hanna
Co-founder
Product-oriented co-founder with prior finance experience, including Balderton Capital.
Founding Story
Armin Ronacher and Colin Daymond Hanna met in Vienna, discovered shared values around software, openness and human agency, and decided after traveling and working together to start a company. They incorporated Earendil as a PBC so that their founding charter—craft software and open protocols, strengthen human agency, bridge division and ignorance, and cultivate lasting joy and understanding—would guide the business. The founders describe beginning in spring 2025 at a Vienna coffee shop without yet having a company or office.
Business Model
Revenue Model
An open-core model: Pi's MIT-licensed core remains free and open source; Earendil plans Fair Source value-add features and proprietary enterprise features/cloud infrastructure to fund the open work. Lefos is currently a public-alpha/waitlist product.
Pricing Tiers
Pi core is MIT licensed and can be used, forked and commercialized by users.
Planned source-available commercial features converting to open source after a delay; not yet built at the time of the announcement.
Planned enterprise-specific features and cloud infrastructure; not yet built at the time of the announcement.
Target Markets
- Individual software developers
- Open-source developers and agent-framework builders
- Engineering teams and enterprises
- Small businesses and companies seeking customizable AI systems
- People and teams using email for collaborative work
- Software engineering and coding-agent workflows
- Building customized AI agent workflows and extensions
- Enterprise engineering experimentation and optimization
- Local-model and multi-provider agent use
- Email-based personal and collaborative work
- Communication, drafting, translation and document analysis
- Databricks
- Shopify
- OpenClaw