Metadata-Version: 2.4
Name: piia-engram
Version: 3.5.0
Summary: Engram — AI 记忆印记。你的 AI 记忆，本地存储，跨工具共享。
Author: Engram Contributors
License: Apache-2.0
Project-URL: Homepage, https://github.com/Patdolitse/engram
Project-URL: Repository, https://github.com/Patdolitse/engram
Project-URL: Bug Tracker, https://github.com/Patdolitse/engram/issues
Keywords: ai,memory,mcp,knowledge,personal-knowledge,llm
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: mcp>=1.0
Requires-Dist: portalocker>=2.0
Provides-Extra: dev
Requires-Dist: pytest>=7.0; extra == "dev"
Requires-Dist: tomli>=2.0; python_version < "3.11" and extra == "dev"
Provides-Extra: remote
Requires-Dist: uvicorn>=0.20; extra == "remote"
Provides-Extra: secure
Requires-Dist: cryptography>=41.0; extra == "secure"
Provides-Extra: all
Requires-Dist: uvicorn>=0.20; extra == "all"
Requires-Dist: cryptography>=41.0; extra == "all"
Dynamic: license-file

<div align="center">

<img src="assets/social_preview.png" alt="Engram - Local AI memory layer" width="640">

# Engram

### A local memory layer for AI coding tools

**Stop re-explaining yourself every time you switch tools, projects, or sessions.**

`Claude Code` | `Codex` | `Cursor` | `MCP compatible` | `100% local`

[English](README.md) | [中文](README.zh-CN.md)

[![License: Apache 2.0](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE)
[![Python 3.10+](https://img.shields.io/badge/Python-3.10%2B-blue.svg)](https://python.org)
[![MCP Compatible](https://img.shields.io/badge/MCP-Compatible-purple.svg)](https://modelcontextprotocol.io)
[![PyPI](https://img.shields.io/pypi/v/piia-engram)](https://pypi.org/project/piia-engram/)

</div>

---

> **TL;DR:** Engram is an MCP server that gives Claude Code, Codex, and Cursor a persistent identity layer — your profile, preferences, lessons learned, and key decisions stored as local JSON files. One write, every AI reads. 100% local, Apache 2.0.

---

AI coding tools are powerful, but they do not really know you.

Every time you switch from Claude Code to Codex, open Cursor, start a new session, or move into a different project, you often have to explain the same things again:

- how you prefer to communicate
- how the AI should approach code
- which project rules matter
- which mistakes should not happen again
- why earlier decisions were made

Engram stores that collaboration memory as local files, then exposes it through MCP so different AI tools can read the same user context.

The goal is simple: **make every compatible AI tool start from the same understanding of you.**

## Why Engram?

| Without Engram | With Engram |
|---|---|
| Every new session starts from zero | AI tools can load your identity and preferences |
| Switching tools loses accumulated context | Claude Code, Codex, and Cursor can read the same memory |
| Project rules live in scattered prompts | Rules and decisions are stored as local assets |
| Past mistakes get repeated | Lessons learned can follow you across tools |
| Memory is locked inside one product | Data stays local, editable, and portable |

Engram is not another chat app, agent framework, or hosted memory service. It is a small local memory layer that sits underneath the tools you already use.

Unlike session-memory tools that remember what happened in a task, Engram stores **who you are** — your identity, preferences, lessons, and decisions — so every AI tool starts from the same understanding of you as a person.

## What It Stores

Engram can store:

- identity and communication preferences
- coding style and quality standards
- trust boundaries for AI tools
- project snapshots
- lessons learned
- key technical or product decisions
- domain knowledge that should be reused later

All data is stored under `~/.engram/` as JSON and Markdown files. You can open, edit, back up, or migrate it yourself.

## Quick Start

```bash
git clone https://github.com/Patdolitse/engram.git
cd engram
pip install piia-engram      # Install from PyPI (recommended)
# Or install from source: pip install -e .
python demos/setup_engram.py
```

Then configure Engram as an MCP server in your AI coding tool.

Example MCP config:

```json
{
  "mcpServers": {
    "engram": {
      "command": "python",
      "args": ["/path/to/engram/src/engram_core/mcp_server.py"]
    }
  }
}
```

After restarting your MCP-compatible client, a new session can call `get_user_context` to understand your profile, preferences, lessons, and project context.

## Remote Deployment

Run Engram on your own server and connect from anywhere.

### Server Setup

```bash
# Install with remote support
pip install piia-engram[remote]

# Generate an auth token
python -c "import secrets; print(secrets.token_urlsafe(32))"
# Save the output, e.g. "abc123..."

# Start in SSE mode
ENGRAM_AUTH_TOKEN=abc123... python -m engram_core.mcp_server --transport sse --host 0.0.0.0 --port 8767
```

### Client Config (Claude Code)

```json
{
  "mcpServers": {
    "engram": {
      "url": "http://your-server:8767/sse",
      "headers": {
        "Authorization": "Bearer abc123..."
      }
    }
  }
}
```

### Client Config (Cursor)

```json
{
  "mcpServers": {
    "engram": {
      "url": "http://your-server:8767/sse",
      "headers": {
        "Authorization": "Bearer abc123..."
      }
    }
  }
}
```

**Security notes:**
- Always use HTTPS in production, behind nginx or caddy with TLS.
- The auth token protects your identity data. Keep it secret.
- Default bind is `127.0.0.1` for localhost only. Use `0.0.0.0` only behind a reverse proxy.
- Data stays on your server and never touches third-party clouds.

## MCP Tools

Engram exposes read, write, project, backup, and compatibility tools through MCP.

Common tools include:

| Tool | Purpose |
|---|---|
| `get_user_context` | Load the complete user context at the start of a session |
| `get_identity_card` | Export a Markdown identity card for tools without MCP |
| `get_profile` | Read the user profile, optionally filtered with `safe=true` |
| `get_preferences` | Read communication and workflow preferences |
| `get_trust_boundaries` | Read data access boundaries |
| `get_quality_standards` | Read quality expectations |
| `get_lessons` | Read reusable lessons learned |
| `get_decisions` | Read key decisions and reasons |
| `get_relevant_knowledge` | Find knowledge relevant to a project |
| `get_knowledge_inheritance` | Build a cross-project knowledge starter pack from free text |
| `save_project_snapshot` | Save project context for later sessions |
| `add_lesson` | Add a lesson learned |
| `add_decision` | Add a key decision |
| `bulk_add_knowledge` | Add multiple lessons or decisions in one call |
| `ingest_notes` | Parse free-form notes into lessons and decisions |
| `extract_session_insights` | Extract lessons and decisions from session summaries |
| `export_engram` | Export a full backup |
| `import_engram` | Import a backup |
| `export_engram_to_openclaw` | Export OpenClaw-compatible files |
| `import_engram_from_openclaw` | Import OpenClaw-compatible files |
| `search_knowledge` | Search lessons and decisions by weighted multi-term relevance |
| `get_knowledge_overview` | Knowledge overview: digest, health report, and stale checks |
| `get_related_knowledge` | Follow links between lessons and decisions |
| `find_similar_knowledge` | Find similar lessons and decisions by content |
| `export_knowledge_report` | Export a readable Markdown knowledge report |
| `link_knowledge` | Create a bidirectional link between two knowledge items |
| `unlink_knowledge` | Remove a bidirectional knowledge link |
| `merge_knowledge` | Merge a duplicate knowledge item into the primary item |
| `update_knowledge` | Update a lesson or decision by ID |
| `archive_knowledge` | Archive a lesson or decision by ID |
| `get_audit_log` | Get recent audit log entries |

## Data Layout

```text
~/.engram/
|-- schema_version.json
|-- identity/
|   |-- profile.json
|   |-- preferences.json
|   |-- quality_standards.json
|   `-- trust_boundaries.json
|-- knowledge/
|   |-- lessons.json
|   |-- decisions.json
|   `-- domains.json
|-- projects/
|   `-- {project_id}.json
|-- exports/
`-- compat/
    `-- openclaw/
```

## Supported Tools

| Tool | Integration | Status |
|---|---|---|
| Claude Code | MCP over stdio | Tested |
| Codex | MCP over stdio | Tested |
| Cursor | MCP over stdio | Expected to work |
| Claude Desktop | MCP over stdio | Expected to work |
| OpenClaw | SOUL.md / MEMORY.md / USER.md import and export | Tested |
| ChatGPT / Gemini / Kimi | Markdown identity card fallback | Usable |

## Comparison

| Feature | Engram | Claude Memory | Manual `CLAUDE.md` | Mem0 |
|---|---|---|---|---|
| Cross-tool sharing | Yes | Claude only | Tool-specific | Yes |
| Local storage | Yes | Cloud | Local | Cloud / hosted |
| Directly editable data | JSON / Markdown | Not visible | Yes | API-based |
| MCP standard | Yes | Not applicable | Not applicable | Yes |
| Portable backup | Copy files or export JSON | Limited | Copy files | API export |
| Model-agnostic | Yes | Claude-focused | Depends on the tool | Yes |
| Price | Free and open source | Included in subscription | Free | Free / paid tiers |

## Built With

Engram is a human-directed, AI-assisted open-source project.

| Contributor | Role |
|---|---|
| [@Patdolitse](https://github.com/Patdolitse) | Creator, product direction, strategy, ownership |
| Claude Code | Architecture, task planning, code review assistance |
| Codex | Implementation, testing, documentation assistance |

## FAQ

**What is Engram?**
Engram is a local-first MCP server that gives AI coding tools (Claude Code, Codex, Cursor) a persistent identity layer. It stores who you are, how you work, what you have learned, and the decisions you have made — as local JSON files on your machine.

**How is Engram different from other AI memory tools?**
Most AI memory tools store what happened in a session (task context, code changes). Engram stores who you are as a person — your identity, preferences, lessons, and decisions. This identity layer persists across tools, sessions, and projects. Your data is local JSON files you own and can edit directly.

**Which AI tools does Engram support?**
Engram works with any MCP-compatible AI tool: Claude Code, OpenAI Codex, Cursor, Claude Desktop, and others. For tools without MCP support (ChatGPT, Gemini, Kimi), you can export a Markdown identity card and paste it in manually.

**How do I install Engram?**
```bash
git clone https://github.com/Patdolitse/engram.git
pip install piia-engram
# Or from source: cd engram && pip install -e .
python demos/setup_engram.py
```
Then add the MCP config and restart your AI tool. The AI will call `get_user_context` automatically at the start of each session.

**Does Engram send data to the cloud?**
All data is stored in `~/.engram/` on your local machine. Engram itself never uploads data anywhere. The optional `read_web_content` tool makes outbound HTTP requests to a local Reader service (`localhost:7890`) which may in turn fetch external URLs — but only when explicitly invoked. Core identity and knowledge tools make no network requests.

**How many MCP tools does Engram provide?**
Engram exposes 37 MCP tools covering identity management, lessons learned, key decisions, project snapshots, bulk input, note ingestion, session insight extraction, weighted knowledge search, similarity discovery, merging, digesting, reporting, linking, health checks, and audit logging.

**Is Engram free?**
Yes. Engram is free and open source under the Apache 2.0 license.

## Limitations

Engram is functional and actively used, but some things it intentionally does not do yet:

| Area | Current State | Planned |
|---|---|---|
| **File safety** | Atomic JSON writes with a shared portalocker file lock | Broader stress testing |
| **Access control** | `restricted_fields` filters profile fields from `get_user_context` and `get_profile(safe=true)` | Per-caller ACL blocked by MCP caller identity |
| **Encryption** | Optional field-level AES-256-GCM encryption via `ENGRAM_SECRET` env var. Install `pip install piia-engram[secure]`. | Full-disk encryption for all files (v4.0) |
| **Audit logging** | Optional access audit log via `ENGRAM_AUDIT=1` env var. Logs to `~/.engram/audit.log`. | Per-caller audit (blocked by MCP spec) |
| **Caller identity** | MCP protocol doesn't pass tool identity | Blocked by MCP spec |
| **Concurrent writes** | Protected by file lock + atomic replace for Engram JSON writes | Network-filesystem edge cases not guaranteed |

**What this means in practice:**
- Don't store passwords, API keys, or client PII in Engram
- Any process with read access to `~/.engram/` can read your data
- `restricted_fields` reduces what Engram emits in cold-start context, but it is not encryption or a true ACL

This is not a warning to avoid Engram — it's an honest description of what it is: a local memory layer for personal AI context. For personal use, it works well today.

## Security Configuration

### Field-level encryption (optional)

Encrypt sensitive profile fields (email, phone, location, etc.) at rest:

```bash
pip install piia-engram[secure]
export ENGRAM_SECRET="your-strong-passphrase"
```

Encrypted fields are stored as `enc:v1:...` in JSON files. Without `ENGRAM_SECRET`, Engram works normally with plaintext (backward compatible).

### Audit logging (optional)

Track all read/write operations:

```bash
export ENGRAM_AUDIT=1
```

Logs are written to `~/.engram/audit.log` in JSON-lines format. Query with `get_audit_log` tool or `grep`.

## Contributing

Contributions, issues, and feedback are welcome.

See [CONTRIBUTING.md](CONTRIBUTING.md). Chinese readers can also use [CONTRIBUTING.zh-CN.md](CONTRIBUTING.zh-CN.md).

## License

[Apache 2.0](LICENSE). Engram is free software. Your memory belongs to you.
