ai-context

KO | EN

gitlite — search

#ai #typescript #llm #go #android #machine-learning #linux #java #nodejs #windows #cli #awesome-list #pytorch #security

(★ 11)

Unified context system for seamless multi-agent AI development. Works with Cursor, Claude Code, Codex, and Google Antigravity. Single source of truth, session continuity, and zero context loss when switching between AI assistants.

ai-context 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# AI Context

**Context infrastructure — the agent harness — for AI coding agents.** Keep `AGENTS.md`, `CLAUDE.md`, and Cursor rules thin, and put durable, versioned project memory in `.ai-context/` — so Claude Code, Cursor, and Codex all work from the same map of decisions, plans, standards, and session history.

One `ai-context` command manages that context, checks it for drift, and compacts old sessions as your project grows.

[![npm](https://img.shields.io/npm/v/@dkothule/ai-context.svg)](https://www.npmjs.com/package/@dkothule/ai-context) [![downloads](https://img.shields.io/npm/dm/@dkothule/ai-context.svg)](https://www.npmjs.com/package/@dkothule/ai-context) [![node](https://img.shields.io/node/v/@dkothule/ai-context.svg)](https://nodejs.org) [![license](https://img.shields.io/badge/license-MIT-blue.svg)](https://raw.githubusercontent.com/dkothule/ai-context/main/LICENSE)

```bash
npm install -g @dkothule/ai-context
ai-context init
```

Install once, run `ai-context init` in a repo, and every supported agent starts from the same project context. File installation is quick; repo-specific setup runs through your chosen agent and can take a few minutes on larger projects.

![ai-context demo — install, init, and the shared .ai-context/ files agents read](https://raw.githubusercontent.com/dkothule/ai-context/main/docs/ai-context-demo.gif)

---

## Why this exists

AI coding agents are getting better at writing code, but they still lose leverage when project context is scattered across chats, local memories, stale READMEs, and one-off instructions. You ask Claude to refactor something. Tomorrow you switch to Cursor. Next week a teammate opens Codex. Unless the repository itself carries the context, every agent starts from zero.

`ai-context` turns context into project infrastructure: versioned, reviewable, local-first files that every agent can read and update.

## What you get

| | |
|---|---|
| 🧠 **Shared memory** | One context directory, read by Claude Code, Cursor, and Codex alike. |
| 🔁 **Session continuity** | Session logs carry decisions, plans, and state across sessions. |
| 🪝 **Survives compaction** | Per-agent hooks checkpoint the transcript before context is dropped, then prompt the next session to curate it. |
| 🔍 **Drift detection** | `check-drift` audits your context against the real codebase and can auto-apply fixes by severity. |
| 🗜️ **Session compaction** | `compact` rolls up stale logs so "read the latest session" stays useful after months of work. |
| ⚙️ **One CLI, three agents** | `setup` / `check-drift` / `compact` run non-interactively through whichever agent CLI you configure. |
| 🔒 **Reversible & local-first** | Installs back up before they write; files stay local unless your selected agent CLI sends context to its provider. |

## The pattern

This project has shipped the pattern since its first open-source commit in **December 2025**: keep `AGENTS.md`, `CLAUDE.md`, and Cursor rules thin, and make a durable, versioned `.ai-context/` directory the system of record that every agent reads from.

OpenAI's [harness engineering post](https://openai.com/index/harness-engineering/) (February 2026) later reached the same conclusion independently — giant instruction files waste scarce context, go stale quickly, and resist verification, so repository knowledge should be the system of record and the injected `AGENTS.md` a map, not a manual. It's encouraging external validation of an approach `ai-context` already automates.

`ai-context` packages that pattern for real repositories:

- **One shared directory** (`.ai-context/`) that acts as the project context system of record.
- **Thin adapter files** (`AGENTS.md`, `CLAUDE.md`, `.cursor/rules/…`) that act as tables of contents, not monoliths.
- **Scoped context files** for overview, structure, standards, tasks, decisions, changelog, plans, and sessions.
- **Session hooks** that remind agents to log work and preserve transcript breadcrumbs before compaction drops context.
- **Maintenance commands** that mechanically check drift, compact old sessions, and keep shared context usable over time.

### Why not just maintain a `CLAUDE.md` by hand?

- **A single adapter file locks you to one agent.** `ai-context` keeps Claude, Cursor, and Codex reading the *same* source through thin per-agent adapters — switch tools without rewriting anything.
- **Hand-maintained docs go stale silently.** `check-drift` finds the drift and patches it, so your context tracks the code instead of slowly lying about it.
- **Flat instructions don't survive compaction.** The hooks checkpoint and restore transcript context automatically — a static file can't.
- **It scales past one repo and one person.** Session logs, decision records, and shared standards give the next contributor — human or agent — a real handoff, not a wall of instructions.

---

## Get started

Run `ai-context init` in any repo (the install command is in the intro above). Interactive prompts pick your agents (Claude, Cursor, Codex — all pre-checked), then an LLM reads your repo and configures the context files. Because it's installed globally, every later command is just `ai-context <cmd>` — `check-drift`, `compact`, `status`, `use` — in any project, no prefix.

> `ai-context` is language-agnostic — it runs on Python, Go, Rust, or any repo, not just Node projects.

> **Just want to try it without installing?** `npx @dkothule/ai-context@latest init` runs the scaffold once and installs nothing — but then every later command needs the `npx …` prefix too, so the global install above is the better default for ongoing use.

### What `init` does

1. **Asks you** which agent adapters to install.
2. **Backs up** existing `.ai-context/`, `CLAUDE.md`, `AGENTS.md`, and any current agent adapters to `.ai-context-backups/<timestamp>/`. Nothing destructive happens without a safety copy.
3. **Copies templates** into your project: generic `.ai-context/` files, thin adapter wrappers (`CLAUDE.md`, `AGENTS.md`, `.cursor/rules/main.mdc`, etc.), and per-agent session-management hooks under `.claude/hooks/`, `.cursor/hooks/`, and `.codex/hooks/`.
4. **Runs a setup prompt** via your configured CLI (`claude -p`, `codex exec`, or `agent --print`). The agent reads your repo and personalizes `project.overview.md`, `project.structure.md`, and any language/testing standards it detects (TypeScript, Python, Go, etc.).
5. **Writes an install log** to `.ai-context/logs/install/<timestamp>.md` so you can see exactly what happened.

After `init`, start a session with any supported agent — it will read `.ai-context/` on its own. No further configuration needed.

### Other ways to install / upgrade

```bash
npm install -g @dkothule/ai-context@latest && ai-context init  # upgrade to latest — safely handles existing installs

ai-context init --agents claude,cursor,codex --yes   # non-interactive (for CI / scripts)
ai-context init --agents codex --yes --cli codex      # non-interactive + run setup via Codex
ai-context apply                     # install/upgrade only (no setup) — fully non-interactive
ai-context init --dry-run            # preview without writing
ai-context init --gitignore          # also add sessions/ + backups to .gitignore
```

`--yes` skips every prompt, including the post-install CLI picker: it runs setup via `--cli` if given, otherwise the first selected agent (and saves that choice as `configured_cli`). For install/upgrade with no setup step at all, use `ai-context apply`.

**Requirements:** Node 18+. Any Unix-like shell for the bash hooks (macOS, Linux, WSL, Git Bash on Windows).

---

## Commands

Each command does one thing:

| Command | What it does |
|---|---|
| `ai-context init` | Install or upgrade. Interactive prompts for agents + setup. |
| `ai-context apply` | Install/upgrade files non-interactively without running setup. |
| `ai-context setup` | Re-run just the setup prompt (your agent configures `.ai-context/`). |
| `ai-context use [cli]` | Set the CLI agent (claude/codex/cursor) used by setup/check-drift/compact. |
| `ai-context check-drift` | Audit `.ai-context/` vs. your actual code. Optionally apply patches. |
| `ai-context compact` | Archive old session logs to reduce noise. |
| `ai-context status` | Show installed version + schema + configured CLI. |
| `ai-context uninstall` | Remove everything. |

All LLM-driven commands (`setup`, `check-drift`, `compact`) run via a coding-agent CLI and fall back to copying the prompt to your clipboard if none is available or authenticated.

**Which CLI runs them?** The CLI you pick during `init`/`setup` is remembered (saved as `configured_cli` in `.ai-context/manifest.json`), and `setup`, `check-drift`, and `compact` use it by default — so a project set up with Codex keeps using Codex instead of silently falling back to Claude. To change it later, run `ai-context use` (interactive picker) or `ai-context use codex` (direct). A one-off `--cli <name>` flag still overrides it per run, and if no CLI is configured the commands auto-detect a local `claude`, `codex`, or Cursor `agent` (with pre-approved permissions for Claude, a `workspace-write` sandbox for Codex, and `--force` for Cursor).

**Getting help:** every command supports `--help` (or `-h`) for full flag reference:

```bash
ai-context --help                    # list all commands
ai-context check-drift --help        # flags for a specific command
ai-context help check-drift          # same — subcommand form
```

---

## What gets installed

```
your-project/
├── .ai-context/                 # shared context (the whole point)
│   ├── project.overview.md      # what this project is
│   ├── project.tasks.md         # what's in flight
│   ├── project.decisions.md     # architecture decisions (ADRs)
│   ├── project.changelog.md     # user-visible changes
│   ├── project.backlog.md       # ideas, deferred work
│   ├── project.structure.md     # directory layout
│   ├── plans/                   # design plans (write before non-trivial work)
│   ├── sessions/                # session logs (written at session end)
│   │   └── _archive/            # rollups of old sessions (don't read at session start)
│   ├── standards/               # coding + workflow rules
│   └── logs/                    # audit log of every ai-context command run
│
├── CLAUDE.md                    # thin wrapper → @AGENTS.md
├── AGENTS.md                    # thin wrapper → points to .ai-context/
├── .cursor/
│   ├── rules/main.mdc           # Cursor adapter
│   ├── hooks/                   # session-management hooks (preCompact, sessionEnd, sessionStart)
│   └── hooks.json               # Cursor hook registrations
├── .codex/
│   ├── hooks/                   # session-management hooks (Stop, PreCompact, PostCompact, SessionStart)
│   ├── hooks.json               # Codex hook registrations
│   └── config.toml              # enables `hooks` feature flag (required by Codex)
├── .claude/
│   ├── hooks/                   # context-preservation hooks (see below)
│   └── settings.json            # hook registrations
└── your source code...
```

**Principle**: adapter files are thin. All real content lives in `.ai-context/`, loaded on demand by each agent. This keeps context windows clean — agents read only what they need for the current task.

---

## How agents use it

On session start, every agent follows the same tiered reading protocol:

**Always read:**
1. `project.overview.md` — project state and objectives
2. `project.changelog.md` — recent user-visible changes
3. Latest file in `sessions/` (excluding `_archive/`) — last session's handoff

**Then read based on task:**
- Writing code → `standards/project.rules.base.md` + `project.rules.md`
- Planning non-trivial work → `project.tasks.md` + `plans/`
- Continuing prior work → additional files in `sessions/`
- Language/testing specifics → relevant files in `standards/`

At session end, the agent writes a session log, updates `project.tasks.md`, logs decisions to `project.decisions.md`, and notes user-visible changes in `project.changelog.md`.

---

## Hooks

Session-management hooks are installed for Claude Code, Cursor, and Codex. Coverage by agent:

| Agent | Session-end log reminder | Pre-compact autosave | Post-compact reminder |
|---|---|---|---|
| **Claude Code** (`.claude/settings.json`) | ✅ `Stop` | ✅ `PreCompact` | ✅ `SessionStart(compact)` |
| **Cursor** (`.cursor/hooks.json`) | ✅ `sessionEnd` | ✅ `preCompact` | ✅ `sessionStart` (`additional_context`) |
| **Codex** (`.codex/hooks.json`) | ✅ `Stop` | ✅ `PreCompact` | ✅ `PostCompact` + `SessionStart` |

What each hook does:

- **Session-end log reminder** — reminds the agent to write a session log if today's log is missing. Advisory only (never blocks).
- **Pre-compact autosave** — writes `sessions/YYYY-MM-DD-HHMM-precompact-autosave.md` before compaction. The autosave includes a `local_transcript_ref` pointer to the local agent transcript JSONL and, when possible, a short recent-turn excerpt, so critical context remains recoverable even if the compacted summary omits it.
- **Post-compact reminder** — in the fresh session after compaction, surfaces the autosave so the agent curates it into a proper session log. The session log should preserve `source_autosave` and the autosave's `local_transcript_ref` when present, then delete the autosave. Older autosaves may use the legacy name `transcript_ref`; copy it into `local_transcript_ref` during curation.

Result: with Claude Code, Cursor, or Codex, compaction writes a recoverable checkpoint before context is dropped and then reminds the next session to curate it.

The local transcript reference is a fallback, not the durable memory itself. It helps the same local agent recover exact prior discussion if something important was lost during compaction; the curated session log remains the portable handoff future agents should read first. Hooks write the path home-relative when possible (`~/.codex/...` instead of `/Users/name/...`), but if you commit or share session logs, still treat `local_transcript_ref` as local/private and redact it if needed.

All hooks are installed automatically by `ai-context init` for the agents you select. Existing user-owned hooks in `hooks.json` / `settings.json` are preserved — the installer merges additively.

> [!IMPORTANT]
> **Codex requires a one-time trust step, or its hooks silently do nothing.**
> Codex loads `.codex/hooks.json` only after you explicitly trust each hook — installing the files is not enough. After `ai-context init` (or any time the Codex hooks change), trust them one of two ways:
>
> - **Codex CLI** — open Codex in the project and run `/hooks`, then approve the AI Context entries (`PreCompact`, `PostCompact`, `Stop`, `SessionStart`).
> - **Codex desktop app** — go to **Settings → Hooks** and trust the project-level hooks.
>
> Until you do this, Codex's autosave-on-compaction and session-log reminders will **not** run. Trust is stored per hook hash in `~/.codex/config.toml`, so re-trust after the hooks change. If a Codex session was already open when the hooks were added, **restart the app/session (or re-open the project)** before trusting them.

**What about Claude Code and Cursor?** They don't need this explicit per-hook approval:

- **Claude Code** loads hooks from `.claude/settings.json` on startup. If the hooks change mid-session it shows a review notice (`/hooks`); otherwise restarting the session is enough — there's no separate trust gate that blocks them.
- **Cursor** loads `.cursor/hooks.json` under normal workspace trust; no per-hook approval. Restart Cursor (or reload the window) if hooks were added while it was open.

In short: **only Codex hard-blocks hooks behind an explicit trust step.** For all three, restarting the agent after install is the safe move.

---

## Drift detection

Over time, `.ai-context/` drifts from your actual code. Files get renamed, architecture evolves, the overview goes stale. `check-drift` catches this:

```bash
ai-context check-drift                # analyze + write report to .ai-context/logs/drift/
ai-context check-drift --static-only  # fast local checks only (no LLM)
ai-context check-drift --copy         # write report + copy follow-up prompt to clipboard
ai-context check-drift --print        # print the analysis prompt to stdout (no file, no execute)
ai-context check-drift --dry-run      # preview what would happen; no files written
```

Two layers:

1. **Static checks** (local, fast): broken refs in `project.structure.md`, stale `last_updated` frontmatter, "In Progress" tasks with no recent commits, backlog items older than 90 days.
2. **LLM analysis** (optional, via your configured CLI): compares overview/structure/decisions against recent git log + tree, produces a structured report with severity-tagged patches (`[significant]`, `[moderate]`, `[minor]`).

Reports land in `.ai-context/logs/drift/<timestamp>-drift.md`.

### Auto-applying patches with `--fix`

`--fix [severity]` runs drift analysis, **then** invokes a second LLM pass that reads the report and applies matching patches. The severity argument is a **cutoff** — patches at that severity and anything more critical are applied:

```bash
ai-context check-drift --fix               # default: only [significant] patches
ai-context check-drift --fix=significant   # same as above — most critical only
ai-context check-drift --fix=moderate      # [significant] + [moderate]
ai-context check-drift --fix=minor         # everything (includes cleanup/polish items)
ai-context check-drift --fix=all           # alias for minor — full YOLO
ai-context check-drift --fix --dry-run     # print what would be patched, don't write anything
```

Tip: snapshot `.ai-context/` first (`cd .ai-context && git add -A && git commit -m "pre-fix"`) if you want an easy rollback. The fix step edits files in place; reviewing a clean diff afterwards is the safety net.

---

## Choosing the CLI agent

`setup`, `check-drift`, and `compact` need a coding-agent CLI to run their LLM prompt. The CLI you pick the first time you run `init`/`setup` is **saved to `.ai-context/manifest.json`** (`configured_cli`) and reused by every later command — so the project stays on the agent you chose instead of always falling back to Claude.

```bash
ai-context use            # interactive picker (claude / codex / cursor), defaults to current
ai-context use codex      # set it directly
ai-context status         # shows "Configured CLI: codex"
```

- The value is preserved across upgrades (`init`/`apply` won't reset it).
- `ai-context use` saves your choice even if that CLI isn't installed/authenticated yet (it warns, then persists) so you can configure it ahead of time.
- A per-run `--cli <name>` flag on `setup`/`check-drift`/`compact` still overrides the saved default.
- If nothing is configured, the commands auto-detect the first available CLI (and copy the prompt to your clipboard if none is ready).

---

## Session compaction

Session logs accumulate. After a few months you'll have dozens; after a year, hundreds. That noise makes "read the latest session at session start" less useful because there's too much "latest" to sort through.

`ai-context compact` summarizes old sessions into a single rollup file and deletes the originals — keeping `sessions/` focused on recent, actionable context without losing long-term history.

```bash
ai-context compact --dry-run                 # preview what would be archived (safe)
ai-context compact                           # archive > 30 days old, keep latest 10
ai-context compact --older-than 90           # only > 90 days old
ai-context compact --keep 20                 # always preserve the newest 20
ai-context compact --older-than 90 --keep 50 # combine
ai-context compact --copy                    # prompt → clipboard, skip CLI execution
ai-context compact --print                   # prompt → stdout, for pipes
ai-context compact --cli codex               # force a specific CLI
```

**What happens:**
1. Selects sessions that are older than `--older-than` days AND outside the latest `--keep`.
2. Builds an LLM prompt listing the selected files + a rollup template.
3. The agent reads each file, extracts decisions / open threads / file knowledge, writes a rollup at `.ai-context/sessions/_archive/YYYY-MM-rollup.md`, deletes the originals.
4. An operation log lands at `.ai-context/logs/compact/<ts>.md`.

**Rollup format** (`sessions/_archive/YYYY-MM-rollup.md`):

```markdown
---
archived: true
range_start: 2026-02-27
range_end: 2026-03-03
source_count: 4
---

# Archived sessions 2026-02-27 → 2026-03-03

## Decisions carried forward
- ADRs 007–011 added for agent architecture; source: 2026-02-27-bootstrap-ai-context.md
- ...

## Open threads at end of range
- Architecture diagram not yet created; source: ...

## File/area knowledge
- knowledge/: runtime content lives under live/, staging area under staging/

## Archived sessions
- 2026-02-27-bootstrap-ai-context.md — one-line summary
- ...
```

`archived: true` in frontmatter + the `_archive/` README tell agents to skip this folder at session start. `grep` still works when you need a specific historical answer.

If the result isn't what you wanted → see **Safety & rollback** below.

---

## Audit trail

Every command that produces meaningful output writes a log to `.ai-context/logs/`:

```
.ai-context/logs/
├── install/       # ai-context init / apply
├── setup/         # ai-context setup
├── drift/         # ai-context check-drift reports
└── compact/       # ai-context compact operations
```

Logs are append-only — past runs aren't overwritten on upgrade. Review them to see what `ai-context` did to your project, when, and what the agent said.

Most adopters gitignore `.ai-context/logs/` since logs are machine-local and noisy.

---

## Safety & rollback

Everything `ai-context` does is reversible. Three scenarios, three recovery paths.

### Before running anything risky — snapshot `.ai-context/`

If you track `.ai-context/` in git (recommended), make a snapshot before any LLM-driven command. This gives you a one-command rollback:

```bash
cd .ai-context
git add -A && git commit -m "snapshot before <command>"
cd ..

# run whatever
ai-context check-drift --fix
# or
ai-context compact
# or
ai-context init
```

If the outcome isn't what you want:

```bash
cd .ai-context
git checkout -- .           # revert tracked changes
git clean -fd               # remove untracked files (rollup, logs, etc.)
```

### `check-drift --fix` edited files you don't like

```bash
cd .ai-context
git diff project.structure.md project.overview.md   # review what changed
git checkout -- project.structure.md                 # revert one file
# or
git checkout -- .                                    # revert everything
```

The original drift report stays in `logs/drift/<ts>-drift.md` — you can re-read what was supposed to change and cherry-pick manually.

### `compact` deleted sessions you wanted back

```bash
cd .ai-context
git checkout -- sessions/                           # restores deleted session files
rm sessions/_archive/<YYYY-MM>-rollup.md            # remove the rollup (it's untracked)
```

The rollup file itself is valuable reading before deletion — it distilled what was in those sessions. Review it first, then decide.

### `init` / `apply` (upgrade) went wrong

Every install backs up the pre-install state automatically. No git required:

```bash
ls .ai-context-backups/                              # list timestamped backups
# 20260418-081533-19339/   ← pick one

TARGET=20260418-081533-19339
# restore (overwrites current state):
rm -rf .ai-context .cursor .codex .claude/hooks AGENTS.md CLAUDE.md
cp -R .ai-context-backups/$TARGET/* .
```

Or just re-run `ai-context init` against any version — the installer handles downgrades and re-applies cleanly.

### Full uninstall

```bash
ai-context uninstall --dry-run                       # preview what will be removed
ai-context uninstall                                 # remove everything ai-context installed
```

Your source code, git history, and non-AI-Context files are never touched by any of these commands.

---

## Customization

### Agent adapters are thin on purpose

`CLAUDE.md` is literally one line (`@AGENTS.md`) plus a few Claude-specific notes. `AGENTS.md` is ~40 lines. Neither duplicates content from `.ai-context/`. If you want to add project-specific rules, edit `.ai-context/standards/project.rules.md` — not the adapters.

### Language/testing standards are created per-project

The installer doesn't ship example `project.python.md` or `project.testing.md`. Instead, `ai-context init`'s setup prompt analyzes your repo and creates the right standards files based on what it finds (TypeScript, Python, Go, etc.). You can always add more by hand later.

### Sessions stay local if you want

```bash
ai-context init --gitignore        # adds .ai-context/sessions/ and backups to .gitignore
```

Many teams keep session logs local (personal) while committing the rest of `.ai-context/`. Your call.

---

## Supported agents

| Agent | Adapter | Hooks | CLI support for `setup`/`check-drift`/`compact` |
|---|---|---|---|
| **Claude Code** | `CLAUDE.md` + `.claude/hooks/` + `settings.json` | Stop, PreCompact, SessionStart(compact) | ✅ `claude -p` (primary) |
| **Cursor** | `.cursor/rules/main.mdc` | preCompact, sessionEnd, sessionStart (`.cursor/hooks.json`) | ✅ `agent --print` (falls back to `cursor-agent` for older installs) |
| **Codex / OpenAI agents** | `AGENTS.md` + `.codex/hooks/` | Stop, PreCompact, PostCompact, SessionStart (`.codex/hooks.json`) | ✅ `codex exec` |
| **GitHub Copilot** | _(not shipped — incompatible with Copilot's auto-review + can't resolve relative links; backlog: generate a self-contained instructions file)_ | N/A | N/A |

All agents read `.ai-context/`. The CLI column affects whether `ai-context setup/check-drift/compact` can execute the LLM prompt directly vs. copy it to your clipboard for manual paste.

---

## FAQ

**Do I need all the agents?**
No. Pick whatever combo you use during `init`. Rest of the flow is identical.

**Can I use this without AI?**
Yes. Session logs become work journals, standards ensure consistency. Works fine for human-only teams.

**How do I handle secrets?**
Never commit them to `.ai-context/`. Reference environment variables in `project.overview.md` — don't store values.

**What if I want to roll back a change?**
Every action is reversible. See the [Safety & rollback](#safety--rollback) section for recovery paths after `check-drift --fix`, `compact`, or `init`/upgrade.

**Does it work on Windows?**
Yes, via Git Bash or WSL. Native PowerShell equivalents are on the backlog.

**What's the upgrade path?**
Re-run `npm install -g @dkothule/ai-context@latest && ai-context init`. The installer detects existing installs, backs up, upgrades the installer-managed files, and preserves everything project-owned (your overview, tasks, decisions, history, custom standards).

**Something went wrong — how do I see what happened?**
Check `.ai-context/logs/install/` for the latest install, `.ai-context/logs/setup/` for agent output, `.ai-context-backups/<timestamp>/` for pre-upgrade state.

---

## Architecture

Deep dive with mermaid diagrams covering every command and flow: **[docs/ARCHITECTURE.md](docs/ARCHITECTURE.md)**.

Short version:

- **Single source of truth**: `.ai-context/` holds all governance. Adapters are thin pointers.
- **Base vs. local standards**: `project.rules.base.md` ships with the tool (upgraded automatically); `project.rules.md` is project-owned (never overwritten).
- **Ownership-based restore**: on upgrade, tool-owned files are replaced, project-owned files are restored from backup by path pattern. Custom files you add are preserved.
- **Session logs are mandatory**: reminded by Claude's `Stop` hook, Cursor's `sessionEnd` hook, and Codex's `Stop` hook; instructed by the adapter files; and built into the Session Start reading habit.
- **Hooks, not humans, preserve compaction context**: Claude Code's `PreCompact`, Cursor's `preCompact`, and Codex's `PreCompact` hooks autosave the transcript before compaction; the corresponding post-compact/session-start hooks remind the next session to curate the autosave.

### Maintainer note: hook source of truth

Hook behavior has two sources, by design:

- **Script behavior** lives at the repo root in `.claude/hooks/*.sh`, `.cursor/hooks/*.sh`, and `.codex/hooks/*.sh`. `scripts/sync-templates.sh` copies those scripts into `packages/cli/src/templates/<agent>/hooks/` before packaging.
- **Hook registration/config generation** lives in TypeScript: `packages/cli/src/core/claudeHooks.ts`, `cursorHooks.ts`, and `codexHooks.ts`. These modules write or merge `.claude/settings.json`, `.cursor/hooks.json`, `.codex/hooks.json`, and `.codex/config.toml`.

`sync-templates.sh` intentionally removes generated config files such as `cursor/hooks.json`, `codex/hooks.json`, and `codex/config.toml` from the package template tree so upgrades never overwrite user-owned hook configuration.

Codex also requires an explicit hook review step after project-level hooks are added or changed. Run `/hooks` inside Codex for the project and trust the AI Context entries; otherwise Codex may load `.codex/hooks.json` but skip untrusted events such as `PreCompact`/`PostCompact`.

---

## On the horizon

Ideas on the near-term roadmap. Directions, not commitments — open an issue if one would unblock you.

- **Skills** — install and manage skills.
- **Windows-native hooks** — drop the Git Bash / WSL dependency with PowerShell or Node equivalents for the bash hooks.
- **Self-contained GitHub Copilot adapter** — a generator command that builds a flat `.github/copilot-instructions.md` by synthesizing `.ai-context/` content, since Copilot can't resolve relative links the way CLAUDE.md / AGENTS.md can.
- **Hook smoke automation** — add live hook-trigger smoke tests for Cursor/Codex when their CLIs expose stable hook test commands.
- **Plugin system** — register custom static drift checks or additional "Read First" files without forking the tool.
- **`ai-context export`** — dump a flattened snapshot of `.ai-context/` (markdown bundle + manifest) for offline sharing, incident tickets, or attaching to bug reports.

---

## Contributing

PRs welcome. The tool should stay small and sharp.

1. Fork, branch off `main`.
2. `cd packages/cli && npm install && npm test` to confirm the test suite passes.
3. Follow standards in `.ai-context/standards/`.
4. Write a session log for non-trivial changes.
5. Open a PR.

For ideas not listed above, open a GitHub issue — discussion before code is welcome.

---

## License

MIT — see [LICENSE](./LICENSE).

ai-context

ai-context 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

ai-context

ai-context 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README