Stele

Shared project memory for AI coding agents across Claude Code, Cursor, Codex, and Copilot

Freemium· Free forever plan (unlimited public projects, 1 private project, monthly assistant allowance); paid tiers for additional private projectsAgents

Visit website →

Best for

Developers and small teams who juggle multiple AI coding agents (Claude Code, Cursor, Codex, Copilot) and want a persistent, shared project memory that survives tool switches and session resets.

Skip if

Solo developers who stay inside one agent and one IDE, or enterprises with strict data-residency rules who cannot wait for the self-hosted edition.

Stele is a hosted shared-memory layer that lets multiple AI coding agents work on the same project without losing context between sessions or tools. It maintains a per-project graph of decisions, tasks, attempted-and-failed solutions, open risks, and lessons learned, which agents read before acting and update as they work. In practice, this means Claude Code, Cursor, Codex, GitHub Copilot, and any MCP-compatible client can pick up where another agent left off, avoid re-litigating design choices, and stop repeating the same failed refactors. Setup is a single-line CLI install that wires Stele into your MCP config; the platform then onboards itself by ingesting the existing codebase and building an initial project graph. A shared task queue lets agents (or humans directing them) claim work, hand off partial progress, and prevent two sessions from stepping on each other. The graph is self-maintaining: entries get flagged as stale when the code moves on, resolved risks retire themselves, and contradictions surface as conflicts rather than silently overwriting. Typical workflows include running one agent for architecture spikes and another for implementation, alternating between a laptop-based Cursor session and a cloud-based Codex run on the same ticket, or letting a Copilot-driven refactor consult the notes an earlier Claude Code session wrote about why a particular abstraction was rejected. Projects are private by default with export and deletion controls, and a self-hosted SQLite-backed edition has been announced for teams that need full data residency.

Editor's take

Stele is targeting a real pain point: agentic coding tools each have their own scratchpad, and switching between them today means re-explaining every architectural decision. Making the memory MCP-native instead of building yet another IDE plugin is the right call. Worth trying on the free tier if you actually rotate between Cursor, Claude Code, and Codex; less urgent if you have already picked one.

— The AI Tool Bible editorial team

Pros

✅ MCP-native, so it works with Claude Code, Cursor, Codex, and Copilot without per-tool adapters
✅ One-line CLI install with automatic onboarding from an existing repo
✅ Shared task queue prevents two agents from duplicating the same work
✅ Graph-based memory captures failed attempts, not just successful ones, so agents stop repeating dead ends
✅ Free tier is genuinely usable for solo developers (unlimited public projects, one private)
✅ Self-hosted SQLite edition on the roadmap for teams that need local data
✅ Export and deletion controls on private project graphs

Cons

⚠️ Currently hosted-only; the promised self-hosted build is not yet shipped
⚠️ Value depends on running multiple agents or switching tools often, less compelling for single-agent users
⚠️ Shared memory across agents can propagate a bad decision if not curated
⚠️ Early-stage product with limited public track record and small community
⚠️ Private project quota on the free tier is tight for anyone with several client repos

Use cases

shared memory across AI coding agentscross-tool project context handofftask queue for multi-agent workflowsagent onboarding from an existing codebasetracking failed refactor attemptscoordinating Claude Code and Cursor on one repoMCP-based agent memory backendarchitectural decision log for AI agents