The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Aictx
Repo-local continuity runtime for AI coding agents. Helps them continue from the same inspectable work state instead of starting cold. Stop onboarding your coding agents like…
Install
pip install aictxInstall command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →AICTX
Stop onboarding your coding agents from scratch every session.
AICTX is a lightweight repo-local continuity runtime for Codex, Claude Code, GitHub Copilot, and other coding agents. It gives each new session the useful facts from the repository — what happened, what failed, what changed, what was decided, and what should happen next — without turning AICTX into an agent framework, task manager, vector database, or cloud memory product.
pip install aictx
aictx install
aictx init
After that, keep working normally. Compatible agents run the lifecycle themselves:
resume -> work -> finalize -> next session resumes
Quickstart · What AICTX writes · Continuity View · Website
Why AICTX exists
Coding agents are strong, but new sessions still start cold:
- they rescan the same files and docs;
- they rediscover decisions already made;
- they repeat failed commands and stale assumptions;
- useful context gets trapped in one vendor's chat history.
AICTX turns that fragile chat-local memory into repo-local operational continuity. The next agent starts with a compact brief instead of guessing from scratch.
The product loop
The user path is intentionally small:
aictx install
aictx init
then let the coding agent work
The agent path is equally small:
aictx resume --repo . --task "fix the parser bug" --json
# agent edits, tests, debugs, and reviews normally
aictx finalize --repo . --status success --summary "Fixed parser recovery and ran focused parser tests" --json
A later session resumes from local repository state instead of hidden chat history.
You: fix the parser bug
Agent: resumes repo-local continuity
Agent: works normally
Agent: finalizes factual evidence for the next session
What the next agent gets
AICTX records operational facts that make the next session useful faster:
| Continuity | Why it helps |
|---|---|
| Active work | Current task, files, risks, and next action. |
| Handoffs | What the previous session actually left behind. |
| Decisions | Project facts the next agent should not re-litigate. |
| Failures | Known broken commands, errors, and resolved patterns. |
| Validation evidence | Commands, tests, files opened/edited, and final summary. |
| Quality signals | Fresh, stale, missing, obsolete, or unverified context warnings. |
Everything is local and inspectable. The default runtime writes under .aictx/ and generated agent-instruction files; no account, daemon, dashboard, or cloud sync service is required.
See it working
Generate an inspectable continuity report at any time:
aictx view --repo .
Default output:
.aictx/reports/continuity-view.md
.aictx/reports/continuity-map.mmd
Check health and repair guidance:
aictx doctor --repo . --json
AICTX is not hidden memory. It is reviewable repo-local continuity with explicit quality signals.
Why it feels different
Similar plugs
Reverse engineer anything with agents, from app behavior down to native binaries.
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype…
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io