What is the difference between AEQI and Pi?

Pi (earendil-works/pi) is a minimal, open-source terminal coding agent: a small system prompt, a handful of tools, lazy-loaded skills, and your own API key. AEQI is the company computer: ten officer agents in named roles, a quest board, shared memory, and a server that runs a company's operations over time.

Can Pi run a company?

Pi runs one coding loop at a terminal. It has no fixed roles, no shared memory across sessions, and no approval step before an action ships — by design, since its whole thesis is a minimal scaffold. Running a company's operations is a different job with different requirements.

Who should choose Pi over AEQI?

A developer who wants a lean, transparent coding agent in their own terminal, on their own API key, with almost nothing between them and the model. Pi is an excellent answer to exactly that question.

AEQI vs Pi: AEQI

Choose Pi if you want a minimal, transparent coding agent at your terminal, on your own API key, writing and fixing code with as little scaffold between you and the model as possible.

Choose AEQI if you want a company that runs itself over weeks and months: a staffed team in ten named roles, a memory that compounds, a live website, and a server of your own, not a single coding session that ends when the terminal closes.

What Pi is

Pi describes itself as an “AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI” (as of September 2026). It is open-source software, released under the MIT license by earendil-works, built around a handful of packages: a coding agent CLI, a runtime engine handling tool execution, a multi-provider model layer supporting several LLM vendors, and a terminal interface with differential rendering. The unit of the product is the agent loop itself: a very small system prompt, a few base tools — read, write, edit, shell — and skills that load only when needed. There is no subscription; you bring your own model API key, and pricing is not addressed beyond that.

Side by side

Question	Pi	AEQI
Who it's for	A developer who wants a minimal coding agent at a terminal	A founder who wants a company staffed with agents in named roles
Where it lives	Your own terminal, on your own machine	A company computer with a live website of its own
Agents and roles	One agent loop with a handful of base tools — no named roles	Ten officer agents in named roles: CEO, CTO, COO, CFO, CMO, CRO, CPO, Chief of Staff, General Counsel, Founder's Associate
Memory	Not published beyond the single session's context	A shared memory linking mail, decisions, and quests
Channels	Not published — a terminal CLI, not a chat surface	One inbox across Email, Slack, Teams, Telegram, and WhatsApp
Hosting and data	Runs on your own machine; you hold the model API key and the data	One hosted Company per customer on its own server in Germany, EU data residency
Pricing	Free and open source (MIT license); you pay only your own model API usage	€50 (Personal, 1 seat), €200 (Startup, 4 seats), €800 (Business, 16 seats) a month; annual pays for 10 months
Trial	Not applicable — free to use under the MIT license	3-day free trial; 33% off Personal's first three months until 31 October 2026
Best for	Writing and fixing code fast, cheaply, and transparently	Running a company's operations over time, with a staffed team and a memory

Pi is the better choice when

The job is a single coding session — fix this bug, write this feature, ship this patch — and you want the smallest possible scaffold between you and the model, with nothing loaded into context that the task at hand does not need.
You want to read and understand every tool your agent has, rather than trust a large framework you did not write and cannot fully audit, so you can always see exactly what runs and why it ran.
You already have a model API key and would rather pay per token, on whichever provider you choose, than a monthly subscription for infrastructure — named roles, shared memory, a hosted server — that a single coding session does not need at all.

AEQI is the better choice when

The job is not one session but an ongoing company — marketing, sales, finance, and product all running at once, in named roles that stay filled after the terminal session ends.
You want a memory that survives across sessions, linking mail, decisions, and quests in one place, so the company does not forget what it decided last month, or why it decided it.
You want a Decisions pane where an agent asks before anything is published, sent, or paid, with a record kept alongside the work, rather than a single loop acting alone with no one to check it.
You want to choose your own model per seat — Claude Code, OpenCode, Hermes, pi, or the AEQI runtime — for each officer role, inside a company structure that also holds memory, decisions, and a website. Pi itself is one of those five options: an officer role can run on it, rather than pi standing in for the company around it.

The difference in one sentence

Pi is a minimal loop for writing code in a terminal, one session at a time; AEQI is the company that loop would need to work for over time — named roles, a memory that compounds, a server of its own, and a website that outlasts any single session.

Start a Company, free for three days →

Last verified 14 September 2026. Competitor facts are quoted from their public pages on that date.