Projectmem

---

🎬 Watch the demo

Full screen-recorded tutorial- watch on YouTube

📚 Docs

Doc	What's in it
TUTORIAL.md	15-minute step-by-step walkthrough — set up projectmem on your own project, watch the lifecycle, see the pre-commit warning fire.
CHANGELOG.md	Release history. Latest: v0.1.4 — the accountable-judgment release: stale-memory detection, decision supersede, precheck snooze, pjm brief, failed-approach surfacing, CLAUDE.md export, dashboard Overview.
Research paper (arXiv:2606.12329)	PROJECTMEM: A Local-First, Event-Sourced Memory and Judgment Layer for AI Coding Agents — the peer-readable version: design, Memory-as-Governance framing, capability comparison, and the 207-event dogfooding study.
LICENSE	MIT

---

The Problem

Every new AI session starts from zero. Claude, Cursor, Aider — they all forget yesterday's decisions, repeat failed debugging attempts, and burn millions of tokens reconstructing context from raw source files.

The model isn't the problem. The architecture is. Stateless models need a memory cortex.

The Solution

projectmem is the local-first memory + judgment layer that sits above your AI tools. It captures every failed attempt, decision, and gotcha — then injects that experience back into future AI sessions. Git tracks what changed. projectmem tracks why it changed, what was tried, and what failed.

Install

pip install projectmem
cd your-project
pjm init

That's it. pjm init installs three git hooks (pre-commit warnings, post-commit classification, post-merge tracking), auto-starts a real-time file watcher, inherits cross-project memory if available, and creates .projectmem/. Capture is active from minute one.

The canonical command is projectmem. A pjm alias is installed for speed.

Why You'll Love It

• Pre-Commit Warnings — pjm precheck warns you before you commit if you're about to repeat a failed approach, modify a high-churn file, or touch an unresolved issue. No other AI tool does this — it requires the memory layer underneath. The warning now lists the dead ends themselves ("What already failed here: ✗ tried CSS contain:layout"), and pjm precheck --snooze 2h silences it politely — the snooze is itself logged, so even the silence is audited.
• Stale-Memory Detection (new in 0.1.4) — other memory tools silently decay or delete old memories; projectmem never deletes. Every decision that cites a file is cross-checked against that file's git history — when the file has moved on, the memory is flagged ("predates 7 commits to auth.py — confirm or supersede") and a human decides. Retire it cleanly with pjm decision "new way" --supersedes <id>: the old event stays in the log, tagged, forever.
• Session-Start Briefing (new in 0.1.4) — pjm brief answers "where was I?" in one screen: active warnings, possibly-stale memories, open issues, recent decisions, stack gotchas, and your prevention score with a week-over-week delta.
• Memory for agents without MCP (new in 0.1.4) — pjm export --claude-md compiles live decisions, gotchas, and a "Do NOT retry — these already failed" list into a marked block in CLAUDE.md (or .cursorrules). Copilot, plain Claude, any agent that reads the file inherits your project's judgment.
• Smart Context Injection — pjm wrap claude (or cursor/aider) injects a token-budgeted memory block into your AI before the session opens. Your AI starts experienced, not blank.
• Provable ROI Score — pjm score outputs a letter grade (A+ → F) backed by concrete numbers — debugging hours saved, tokens prevented, dollars protected. CI-friendly JSON output and shields.io badge for your README.
• Cross-Project Memory — Lessons learned in one repo follow you forever. Library gotchas, decisions, and patterns live in ~/.projectmem/global/ and auto-inherit into every new project that matches your stack.
• Real-time File Watcher — Background daemon detects rapid edits to the same file (debugging sessions) between commits. Battery-aware, gitignore-aware, auto-started by pjm init.
• Native MCP Server — Plugs into Claude Desktop, Cursor, Antigravity, Codex, and any MCP-compatible tool. 14 native tools force the AI to read context, check files for known failures, and log work automatically. Verified end-to-end against all four clients.
• Interactive Dashboard — pjm visualize opens a four-tab D3.js dashboard: Story Map (failure heatmap), ROI Dashboard, Project Map (tree or graph view), Timeline.
• 100% Local — No cloud, no telemetry, no accounts. Your code, your memory, your machine.

How It Compares

Capability	projectmem	claude-mem	agentmemory	mem0	Letta (MemGPT)
Core focus	Memory + Judgment	Session capture	Memory engine	Chat memory	Agent framework
Pre-commit failure warnings	✅ unique	❌	❌	❌	❌
Stale memory: flag, never delete	✅ new in 0.1.4	❌	❌ silent decay	❌	❌
Supersede without losing history	✅ new in 0.1.4	❌	❌	❌	❌
Captures development history	✅ typed events	🟡	🟡	🟡	🟡
Records architectural decisions	✅	❌	🟡	❌	❌
Memory for agents without MCP	✅ CLAUDE.md export	❌	❌	❌	🟡
Cross-project memory	✅ library-scoped	🟡	🟡	🟡	🟡
Provable ROI score	✅ A+ → F + $	❌	❌	❌	❌
Plain-text, greppable store	✅ events.jsonl	❌	❌	❌	🟡
No server / DB to run	✅ stdio + files	❌	❌	❌	❌ server + DB
No telemetry, no accounts	✅	❌ default-on	✅	❌	🟡
Native MCP server	✅ 14 focused tools	✅	🟡 53 tools	🟡	🟡
Price	✅ Free · MIT	Free + paid tier	Free	Freemium	Free + cloud

_{✅ yes · 🟡 partial · ❌ no — snapshot June 2026; design capabilities, not benchmark results. claude-mem runs a background worker (port 37777) and enables telemetry by default (v13.5+); agentmemory down-ranks and prunes old memories via decay, mem0 rewrites facts on update, Letta's memory blocks self-edit in place — projectmem never deletes: it flags staleness and lets you decide. Letta requires a running server (Postgres or cloud).}

How AI Reads Your Memory (Token Efficiency)

The architecture is built around one rule: AI reads small, distilled files. Tools generate them from the big raw log.

Access mode	Tokens / session	How it works
No projectmem (baseline)	5,000 – 20,000+	AI re-reads source files every session
Universal Mode (markdown)	~2,500	AI reads 3 small distilled files once
MCP Mode (recommended)	~800 – 1,500	AI calls get_summary(), then get_issue(id) only when relevant
pjm wrap (pre-injection)	500 – 2,000	Pre-generated, you set the budget

AI never reads events.jsonl directly. That file is for tools (pjm score, pjm context, pjm wrap). Tools distill the raw log into compact AI-readable summaries.

MCP Integration (Recommended)

Claude Desktop

Easiest — open the config from the UI:

• macOS: Claude menu → Settings… → Developer tab → Local MCP servers → Edit Config.
• Windows / Linux: same path expected (Settings → Developer → Edit Config) — open an issue if your platform differs and we'll update this.

If you prefer the raw file path: ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows.

Paste this block:

"mcpServers": {
  "projectmem": {
    "command": "/opt/anaconda3/bin/python",
    "args": [
      "-m", "projectmem.mcp_server",
      "--root", "/absolute/path/to/your/project"
    ]
  }
}

Two things to know about this block:

• Use the absolute path to python (e.g. /opt/anaconda3/bin/python, or run which python to find yours). Claude Desktop subprocesses don't inherit your shell PATH, so bare "python" often fails.
• We pass the project root via --root, not the cwd JSON field. Claude Desktop's current build (with the Epitaxy / Cowork workspace system) silently ignores the cwd field — the server ends up running with cwd=/ and can't find .projectmem/. The --root flag is honored by projectmem directly (read from sys.argv) and works regardless of how Claude Desktop spawns the subprocess.

Then fully quit Claude Desktop (Cmd+Q on Mac) and reopen — MCP servers only initialize on cold start.

Cursor

Two ways to register the MCP server — pick whichever fits your workflow:

1. Global (recommended): Cursor menu → Settings… → left sidebar Tools & MCPs → Installed MCP Servers → Add Custom MCP. Paste the JSON below.
2. Per-project: drop the JSON into <project-root>/.cursor/mcp.json — only active when that project is open.

{
  "mcpServers": {
    "projectmem": {
      "command": "/opt/anaconda3/bin/python",
      "args": [
        "-m", "projectmem.mcp_server",
        "--root", "/absolute/path/to/your/project"
      ]
    }
  }
}

Two things to know about this block (same gotchas as Claude Desktop):

• Use the absolute path to python (run which python to find yours). Cursor subprocesses don't reliably inherit your shell PATH.
• Pass the project root via --root, not the cwd JSON field. Cursor — like Claude Desktop — silently ignores cwd: the server ends up running with cwd=~ and can't find .projectmem/. The --root flag is honored by projectmem directly and works around the bug.

Then fully quit Cursor (Cmd+Q on Mac) and reopen. projectmem also auto-discovers .projectmem/ by walking up from CWD (like git does for .git/), and honors PROJECTMEM_ROOT and a --root <path> CLI argument.

Antigravity

Antigravity (Google's AI IDE) speaks standard MCP.

Easiest — open the config from the UI:

1. Open the Agent window (the chat panel on the right).
2. Click the ⋯ Additional Options button in the panel header.
3. Choose MCP Servers → Manage MCP Servers → Add new (or Edit Config).

The raw file is at ~/.gemini/antigravity/mcp_config.json if you prefer editing it directly.

Paste this block:

{
  "mcpServers": {
    "projectmem": {
      "command": "python",
      "args": ["-m", "projectmem.mcp_server"],
      "cwd": "/absolute/path/to/your/project"
    }
  }
}

Then fully quit Antigravity (Cmd+Q on Mac) and reopen — MCP servers only initialize on cold start. All 14 projectmem tools register identically to Claude Desktop / Cursor.

Codex

Codex stores MCP config as TOML (not JSON) in ~/.codex/config.toml. There's a UI form at Settings → MCP Servers → Add MCP Server, but during cross-client verification the form's Save button didn't reliably persist — the file-edit path is faster and more reliable.

Easiest — edit ~/.codex/config.toml directly:

Append this block (preserves any existing config):

[mcp_servers.projectmem]
command = "/opt/anaconda3/bin/python"
args = ["-m", "projectmem.mcp_server", "--root", "/absolute/path/to/your/project"]
cwd = "/absolute/path/to/your/project"

Three things to know about this block:

• Use the absolute path to python (run which python to find yours). Codex subprocesses don't reliably inherit your shell PATH.
• Pass the project root via --root in args (defense in depth). The cwd field appears to work in Codex, unlike Claude Desktop and Cursor — but --root costs nothing and saves us if any future Codex build regresses.
• Set your reasoning effort to medium or higher. On low-reasoning Codex skips get_instructions from the session-start trio, which can cause the AI to miss the Setup Mode workflow rules. Medium+ honors the full trio automatically.

Validate the TOML:

python -c "import tomllib; tomllib.load(open('/Users/<you>/.codex/config.toml','rb')); print('OK')"

Should print OK. If not, the parser tells you the offending line.

Then fully quit Codex (Cmd+Q on Mac) and reopen. Same cold-start rule as every other MCP client. Codex MCP servers spawn lazily on the first tool call in a chat session — if you don't see the process in ps aux right after reopening, send any message to a Codex chat and check again.

Reasoning-effort note: Codex's mode selector is at the bottom of the chat input. Set it to medium (not low) for the full session-start trio behavior. Once set, it persists per-session.

First-run permission prompts

On first use in any MCP-capable client (Claude Desktop, Cursor, Antigravity, Codex), your AI will ask permission before each projectmem tool call. This is expected security behavior — MCP clients require explicit consent for every new tool. Approve each tool once and the prompt won't reappear for that session.

Other MCP Tools

Any MCP-compatible client works — point your tool at python -m projectmem.mcp_server and either set cwd to your project root or rely on the parent-walk auto-discovery.

MCP Tools Exposed

All 14 tools your AI can call:

Read-side (9 tools):

Tool	When to use
get_instructions()	Start of every session — load workflow rules
get_summary()	Start and end — distilled project memory
get_project_map()	Start — understand repo structure
precheck_file(path)	Before editing any file — surface failure history
get_issue(id)	Read one specific issue's full history by ID
search_events(query)	Plain-text search across all logged events
get_context(tokens, focus)	Token-budgeted memory block with optional focus filter
get_score()	A+→F prevention score + ROI numbers
get_global_gotchas(library)	Cross-project library lessons inherited from past repos

Write-side (5 tools):

Tool	When to use
log_issue(summary, location)	Immediately when encountering a bug
record_attempt(summary, outcome)	Immediately after each fix attempt (outcome: failed/partial/worked)
record_fix(summary)	After confirming a fix resolves the issue
add_decision(summary, supersedes?)	When making architectural / design decisions; pass supersedes to retire a stale decision without losing history
add_note(summary)	When discovering gotchas, setup details, or constraints

CLI Reference

Core memory

Command	Purpose
pjm init	Initialize memory + auto-install hooks + inherit global memory
pjm log <text>	Start a new issue / debugging session
pjm attempt <text> [--failed|--worked]	Record a fix attempt outcome
pjm fix <text>	Record the confirmed fix and close the issue
pjm decision <text> [--supersedes <id>]	Record an architectural decision; optionally retire a prior one (old event stays in the log, tagged)
pjm note <text>	Record durable context or a gotcha
pjm show	Print the current summary
pjm search <query> [--failed-only]	Plain-text search across all events; --failed-only lists the project's dead ends
pjm brief	One-screen session-start briefing: warnings, stale memories, open issues, decisions, score
pjm export [--claude-md|--cursor]	Compile live memory into CLAUDE.md / .cursorrules for agents without MCP

Intelligence layer

Command	Purpose
pjm watch [--daemon|--stop|--status]	Real-time file churn watcher
pjm precheck [--snooze 2h|--unsnooze]	Warn about repeating failed approaches before commit; snooze politely (audited) when needed
pjm wrap <agent>	Inject token-budgeted memory into Claude/Cursor/Aider
pjm context [--tokens N]	Generate token-budgeted project context
pjm score [--format text|json|badge]	Letter-grade prevention score
pjm global <action>	Manage cross-project memory

Visualization & utility

Command	Purpose
pjm visualize	Open interactive D3.js dashboard
pjm stats	Token ROI summary in the terminal
pjm backfill	Auto-populate memory from git history
pjm hooks install|uninstall	Manage git hooks manually
pjm regenerate	Rebuild summary.md from events.jsonl

Use --at "file.py:42" with any logging command to attach precise location metadata.

Example: Pre-Commit Warnings in Action

$ git commit -m "switch auth to JWT"

projectmem: Pre-Commit Check
─────────────────────────────────────────────
  src/auth/middleware.py
    WARN  What already failed here (2 attempts):
           ✗ tried switching to JWT middleware (2d ago)
           ✗ patched session timeout to 60min (5d ago)
    WARN  HIGH CHURN: 5 changes in last 30 days
    WARN  1 possibly-stale memory cites this file
           decision [evt_9db5a3f8…] "auth uses session
           cookies, 30min timeout" — predates 7 commits
           Confirm it still holds, or retire it:
           pjm decision "..." --supersedes <id>
─────────────────────────────────────────────
3 warning(s). Review before committing.

~30 min re-debugging just saved.

Need it quiet for a refactor sprint? pjm precheck --snooze 2h — warnings pause, the pause itself is logged, and every commit shows one dim line so silence is never mistaken for a clean check.

Privacy & Security

By default, projectmem commits the distilled files (summary.md, PROJECT_MAP.md, AI_INSTRUCTIONS.md, issues/) and gitignores the raw log + runtime files (events.jsonl, watch.pid, watch.log). This means your teammate's AI inherits your team's knowledge automatically — just git clone and the AI already knows what your team learned.

Want total privacy? Add a single line .projectmem/ to your .gitignore. Nothing leaves your machine.

Full security policy and threat model: SECURITY.md · Privacy & Security guide

Design Principles

• Local-first — No network calls, no cloud, no telemetry. Your data never leaves your machine.
• Project-scoped — Memory lives in the repo. When the code moves, the memory moves.
• AI-tool-agnostic — Works natively via MCP, or universally via Markdown instructions. Any AI tool, any workflow.

Built With

projectmem stands on the shoulders of these excellent open-source projects:

• Typer — the CLI framework that makes pjm feel ergonomic
• Model Context Protocol — Anthropic's open spec that lets AI agents talk to local tools
• watchdog — cross-platform filesystem event monitoring (the heart of pjm watch)
• D3.js — the interactive visualizations in pjm visualize

Research & Citation

projectmem is described in a peer-readable research paper:

PROJECTMEM: A Local-First, Event-Sourced Memory and Judgment Layer for AI Coding Agents Ripon Chandra Malo, Tong Qiu — University of Utah arXiv:2606.12329 · cs.SE (cross-list cs.AI)

The paper introduces the Memory-as-Governance framing — memory that doesn't merely answer the agent but acts on its next action — and reports the design, the deterministic pre-commit judgment gate, a capability comparison against 12 contemporary memory systems, and a two-month, 207-event dogfooding study across 10 real projects.

If projectmem is useful in your research or writing, please cite:

@misc{malo2026projectmem,
  title         = {PROJECTMEM: A Local-First, Event-Sourced Memory and
                   Judgment Layer for AI Coding Agents},
  author        = {Malo, Ripon Chandra and Qiu, Tong},
  year          = {2026},
  eprint        = {2606.12329},
  archivePrefix = {arXiv},
  primaryClass  = {cs.SE},
  url           = {https://arxiv.org/abs/2606.12329}
}

License

MIT — free for personal, commercial, and enterprise use forever.

---

Help Us Reach More Developers

We don't need money. We need you.

projectmem is built by one developer for the open-source community. Every star, every share, and every contribution helps the project survive and grow.

• Star the repo — takes one click, helps massively with discovery
• Share on X / LinkedIn — tell other devs they don't have to keep paying AI to relearn their codebase
• Open an issue — bug, feature request, or just feedback
• Contribute code — PRs welcome, see contributing guide
• Using projectmem at work or in a commercial product? Reach out to support@projectmem.dev so we know who's shipping with us. It's free — we just love hearing about it.

Stars and shares matter more than money — but if you really want to: sponsor on GitHub →

---