aigent-os
by Community
Claude Code harness that refreshes its own context instead of compacting: at a threshold it writes a verified capsule, runs /clear, and resumes in a fresh session. Memory is plain
OSS
aigent-os
Added 16 Oct 2026
Overview
A Claude Code harness that refreshes its own context instead of compacting. When a session reaches a threshold, it writes a verified capsule in plain markdown, runs /clear, and resumes in a fresh session. Memory is stored in a local git repository, making history readable via git log. MIT licensed, runs on Node 18+, and includes a Codex adapter.
Best for
Best for
Developers running long Claude Code sessions who want git-backed, auditable context persistence
Use cases
- Restart long-running Claude Code sessions without losing context
- Audit agent decision history through git log
- Run the same memory harness with Codex
Notes
24 stars on GitHub. Last updated 2026-10-13. Licensed MIT.
Pros
- Avoids context compaction by using fresh sessions
- Memory is human-readable markdown in git
- Includes adapters for both Claude Code and Codex
Cons
- Requires Node 18+
- Small user base with only 24 stars
- Limited to supported harnesses (Claude Code, Codex)
Get the free Developer’s Field Guide
A 27-page field guide to the AI coding workflow with Claude. Claude Code, MCP servers, the prompt patterns that work, and what to delegate. Free.
Enter your work email. We send it straight over, plus a few short notes worth knowing. Unsubscribe any time.