- type
- Coding agent
- interface
- Terminal (CLI; installs via `npm install -g @letta-ai/letta-code`)
- openSource
- Yes (Apache-2.0)
- lineage
- Memory-first coding agent built on Letta's own stateful-agents SDK; agents persist learned codebase knowledge across sessions instead of starting fresh each run; Letta's own blog post (self-reported, not independently verified against the public Terminal-Bench leaderboard) claims Letta Code is the number-one model-agnostic open-source coding harness on Terminal-Bench, a narrower category-specific claim distinct from Factory Droid's separately tracked overall '#1 Terminal-Bench (58.75%)' figure
- note
- Vendor blog post is dated Dec 16, 2025, and independent coverage (TechCrunch/Letta background pieces) corroborates a December 2025 launch, not the April 2026 date this round's brief assumed; flagging the discrepancy as a correction rather than a missed search. A separate, earlier Letta blog post (letta.com/blog/terminal-bench, dated Aug 5, 2025, predating the Letta Code product announcement) cites a 42.5% overall score, 4th overall and 2nd among Claude 4 Sonnet agents on Terminal-Bench, built in under 200 lines via Letta's stateful-agents SDK. It is not confirmed whether that scored run is the Letta Code product itself or an earlier SDK integration (the letta-ai/letta-terminalbench repo), so those specific figures are not carried into this record; only the ranking claim taken directly from the Letta Code product page itself is used here. That claim is on a different scope/leaderboard entry than Factory Droid's already-tracked '#1 Terminal-Bench (58.75%)' line; the two are not being compared as the same measurement. GitHub API confirms letta-ai/letta-code is Apache License 2.0, checked 2026-09-07.
- iconUrl
- /research/harness-icons/letta-code.png
