# Stele fact sheet

Updated October 10, 2026. Check [stele-ai.dev](https://stele-ai.dev/) before publication if a detail matters to the story.

| | |
| --- | --- |
| What it is | Shared memory for AI coding agents and the people who work with them |
| Works with | Claude Code, Cursor, Codex and any MCP client |
| Made of | A hosted record, an agent plugin, a command-line tool and a web app |
| Founder | Serkan Yersen |
| Pricing | Free to start. Indie $5.99 a month, Pro $12 a month, Team $20 per person a month |
| Public projects | Free, and readable by anyone at [app.stele-ai.dev/explore](https://app.stele-ai.dev/explore) |
| Website | [stele-ai.dev](https://stele-ai.dev/) |
| Product | [app.stele-ai.dev](https://app.stele-ai.dev/) |
| Social | [@stele_dev on X](https://x.com/stele_dev), [github.com/Stele-Dev](https://github.com/Stele-Dev), [LinkedIn](https://www.linkedin.com/company/stele-ai/) |
| Press contact | [serkan@stele-ai.dev](mailto:serkan@stele-ai.dev) |

## The name

**Stele** is pronounced STEE-lee (/ˈstiː.li/). A stele is an upright stone slab carved with a record that was meant to outlast the people who wrote it. Stele keeps that kind of record for software projects: what was decided, what was learned, and what is still being worked on.

## One line (24 words)

Stele gives software teams and their AI coding agents a shared record of project knowledge and work, so context carries across sessions and tools.

## Short (57 words)

Stele is persistent project memory for people and AI coding agents. It keeps decisions, lessons, documents and tasks linked in one shared record. When an agent starts work, Stele brings forward the relevant history, including what the team tried before and why it changed direction. People can read and steer the same record in the web app.

## Long (111 words)

AI coding agents start every session with no memory of the project. They repeat mistakes the team already fixed and undo decisions they never heard about. Stele gives them a shared record to read first.

Agents in Claude Code, Cursor, Codex and other MCP clients read what the project decided, what is in progress and what went wrong before, then write back what they learn. Tasks and knowledge live in one linked graph, so a new session or a different tool picks up where the last one stopped. People see the same record in a web app, where they can correct it, assign work and publish it as a public project.

## Founder

Serkan Yersen is a software engineer with more than 19 years of experience. At Meta he worked on WhatsApp Web, release infrastructure and developer AI workflows. Before that he was a founding engineer at JotForm. He started Stele after watching agent sessions start cold, relearn the project and rebuild things the team had already tried and thrown out. His account is on the [About page](https://stele-ai.dev/about), and more about him is at [serkan.dev](https://serkan.dev).