Stele
Stele is shared, persistent memory for software projects built with AI coding agents. It keeps decisions, lessons, documents and tasks linked in one project record that both people and agents can read and update across sessions and tools.
At a Glance
- Solo software developers
- Small engineering teams
- Open-source project maintainers and contributors
- Organizations building software with multiple AI coding agents
AI Tools by Stele
(1)
Stele
Shared Project Memory for AI Agents
Discussions
No discussions yet
Be the first to start a discussion about Stele
Latest News
Stele published its press kit and fact sheet, describing the product as shared memory for AI coding agents and confirming open-beta pricing.
Stele AI accepted into the NVIDIA Inception program.
Why we chose OpenTUI over Ink for the Stele CLI.
Published AI agent memory: what it is, the kinds that exist, and how to choose.
Products & Services
A hosted, linked record/knowledge graph for project decisions, lessons, risks, documents, tasks and components, with full-text and semantic retrieval and lifecycle handling for superseded or retired facts.
Agent-facing integrations that let coding agents read relevant project history before acting and write back what they learn.
A single-binary command-line tool that installs and wires Stele into projects and supported coding-agent environments.
A web interface for viewing the project graph, tasks, knowledge, documents and objectives, plus an assistant that answers questions from the live graph and cites the underlying records.
Market Position
Stele positions itself as a shared, hosted project record for teams using multiple AI coding agents, rather than per-agent memory, manually maintained repository Markdown, a general wiki, or a memory API embedded in an application. Its differentiators are typed decisions/tasks/lessons, linked project history, lifecycle-aware truth maintenance and atomic task coordination. Its stated trade-offs are that it is hosted rather than self-hosted and asks agents to write deliberate facts instead of automatically capturing everything. Competitors and alternatives identified in Stele's comparison include built-in agent files/memory, claude-mem, agentmemory, Basic Memory, Mem0, Supermemory, Zep/Graphiti, Cognee, ByteRover and Letta Code.
Leadership
Founders
Serkan Yersen
Founder and engineer with about 20 years of software experience. He spent nearly seven years at Meta working on WhatsApp Web, release infrastructure and agentic-AI workflows; before Meta he was a founding engineer at JotForm, and he was a frontend architect at Redbeacon/TechShed supporting Home Depot Pro workflows.
Executive Team
Serkan Yersen
Founder & Engineer
Former Meta staff software engineer focused on WhatsApp Web, release infrastructure and agentic-AI workflows; former JotForm founding engineer and Redbeacon/TechShed frontend architect.
Founding Story
Serkan Yersen started Stele after seeing AI-agent sessions begin cold, repeatedly relearn projects and rebuild approaches that teams had already tried and discarded. His initial vision was a durable project record containing decisions, dead ends, lessons and work so every person and agent could start from what was already known.
Business Model
Revenue Model
Subscription SaaS with a permanently free tier and paid monthly plans differentiated by private-project, knowledge-write, storage, assistant, membership and support limits. Enterprise offers custom limits, SLA and add-ons.
Pricing Tiers
Unlimited public projects, 1 private project, 500 new knowledge writes/month, 5,000 stored nodes and 250 assistant turns/month.
3 private projects, 1,500 knowledge writes/month, 15,000 stored nodes, 500 assistant turns/month and email support.
Unlimited private projects, 3,000 knowledge writes/month, 50,000 stored nodes, up to 5 members per private project and 1,500 assistant turns/month.
Unlimited members and projects, 5,000 pooled assistant turns/month, roles, audit log, cross-project search and dedicated support; agents are free.
Large repositories, custom limits, SLA and relationship; SSO and self-hosting are add-ons from Team upward.
Target Markets
- Solo software developers
- Small engineering teams
- Open-source project maintainers and contributors
- Organizations building software with multiple AI coding agents
- Solo developers using AI coding agents
- Small software teams sharing project context across agents
- Switching among Claude Code, Cursor, Codex and other agent tools without losing history
- Preventing agents from repeating reverted approaches and known mistakes
- Project onboarding and maintaining institutional memory
- Coordinating concurrent human and agent tasks