MIT-licensed tools reduce coding agent context waste without secondary LLMs
bonsaisushi · reddit · 2026-08-28
The author released two open-source, MIT-licensed tools to address context waste in coding agents without introducing another LLM into the loop.
- Sando: Handles unnecessary context during a session. It locally redacts secrets, caps oversized tool results, and trims request history before transmission, avoiding summarizer models or secondary LLM calls.
- session-handoff: Manages session lifetime. It extracts the working state needed to continue a task in a fresh Claude Code or Codex session and supports migrating active sessions between the two clients.
These tools aim to reduce live agent-generated context and preserve task state when restarting is cheaper than carrying the session forward. The packages reached 1,500 combined downloads in the first 48 hours.
More from coding & agent
- Anthropic tests Hub Mode in Claude for sub-agent management — testingcatalog · 2026-08-28
- Anthropic tests 'Hub' feature for task orchestration in Claude — testingcatalog · 2026-08-28
- Weak RL may trigger communication traits hidden in pretraining — xuanalogue · 2026-08-28
- Tool call dispositions may generalize to unsanctioned agent communication — xuanalogue · 2026-08-28
- OpenAI agent message-leaving may emerge from pretraining priors — xuanalogue · 2026-08-28
- TraceDiff prototype debugs voice agents by diffing success vs. failure traces — Responsible-Dot8405 · 2026-08-28