aictx — Cross-AI Session Context

How it works

THREE COMMANDS.
FULL CONTEXT.

Init once. Save at end of every session. Resume any time — from any AI tool.

Step 01 / Init

ONE-TIME SETUP

Run once in your project directory. Creates .ai-context/context.db — a portable SQLite file that travels with your project.

$ aictx init

Step 02 / Save

END OF SESSION

Summarize what was done. Add todos. Record decisions. Export to markdown. One command or use the /session-save skill.

$ aictx save "what was done"

Step 03 / Resume

START OF SESSION

Any AI reads .ai-context/current.md or you run /session-resume in Claude Code. Full briefing in seconds.

$ aictx resume

CLI Reference

EVERY COMMAND,
DEMOED.

aictx init Initialize project

aictx save Save session snapshot

aictx todo Manage tasks

aictx resume Print context for AI

aictx decide Record key decision

aictx watch Live TUI dashboard

aictx log Session history

$ aictx init initialized .ai-context/ in /your-project project: your-project tip: git add .ai-context/current.md
$ aictx set url "https://github.com/org/repo" set url = https://github.com/org/repo

$ AI_TOOL=claude aictx save "added auth, fixed DB migration, next: write tests" session #4 saved exported → .ai-context/current.md

$ aictx todo add "write integration tests" 2 todo #6 added: write integration tests
$ aictx todo list ● [#3] (in_progress) implement auth middleware ○ [#6] (open) write integration tests ✔ [#1] (done) set up postgres schema

=== AI CONTEXT: talent-scraper ===
LAST SESSIONS: [May 30 21:40][claude] Fixed Makefile, added LeetCode scraper, brainstormed startup discovery
IN PROGRESS: >> implement auth middleware
OPEN TASKS: [#6] write integration tests [#7] deploy to staging
KEY DECISIONS: * Use JWT not sessions — stateless requirement

$ aictx decide "Use JWT not sessions" "stateless — scales horizontally" decision recorded: Use JWT not sessions exported → .ai-context/current.md

┌─ aictx ─────────────── talent-scraper ─ 21:44:02 ─┐ │ │ │ ● IN PROGRESS │ │ [#3] implement auth middleware │ │ ○ OPEN │ │ [#6] write integration tests │ │ [#7] deploy to staging │ │ │ │ open:2 active:1 done:3 │ sessions:8 │ └──────────────────── q=quit r=refresh ─────────────┘

$ aictx log 4 last 4 sessions:
[May 30 21:40] [claude] Fixed Makefile, replaced CodeChef→LeetCode [May 30 18:20] [cursor] Added todo endpoints to gRPC API [May 29 14:05] [claude] Initial DB schema + migrations [May 28 11:30] [gemini] Scaffolded project structure

Cross-AI Portability

ONE FILE,
EVERY AI.

Every save exports .ai-context/current.md — plain markdown any AI can read without the CLI.

.ai-context/current.md

raw markdown # AI Context

**Project:** talent-scraper
**URL:** https://github.com/org/repo
**Updated:** 2026-05-30 21:40

## In Progress

- [ ] implement auth middleware

## Open Tasks

- [ ] [#6] write integration tests
- [ ] [#7] deploy to staging

## Key Decisions

- **Use JWT not sessions** — stateless

## Recent Sessions

**[May 30 21:40] claude:** Fixed Makefile,
added LeetCode scraper

rendered

AI Context

Project: talent-scraper · Updated: 2026-05-30 21:40

In Progress

implement auth middleware

Open Tasks

[#6] write integration tests

[#7] deploy to staging

Key Decisions

Use JWT not sessions — stateless

Recent Sessions

[May 30 21:40] claude
Fixed Makefile, added LeetCode scraper

Integration

PASTE. PROMPT.
CONTINUE.

CLAUDE

Native skills. /session-resume loads context automatically at session start.

/session-resume
/session-save

CURSOR

Reference .ai-context/current.md in .cursorrules for auto-load on project open.

// .cursorrules
Read .ai-context/current.md

GEMINI

Pipe the markdown file directly into Gemini CLI before starting work.

gemini < .ai-context/current.md

CHATGPT

Copy and paste current.md at the start of any new conversation thread.

cat .ai-context/current.md
# paste into chat

NEVERSTARTCOLD

NEVER
START
COLD