Context7
by upstash
Feeds AI the latest docs for libraries and frameworks before it answers, so code it writes does not use outdated, removed functions.
Computer or cloud★ 64,224MITUpdated today
Install
darwin-memo
Overview
Description from the maker (original English)
Self-curating agent memory kept honest by survival-based selection over measured outcomes.
PopNx MCP’s three pillars
These three are never merged into one score: opinions, measurements and popularity are different things.
- User voiceiNo reviews yetReviews open together with accounts in a later step
- Our basic checksi1 of 1 passedChecked from public information; code not scanned yet
- UsageiNot measured yet★ –
What you can do with it
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
- Connection: Runs on your computer
- Transport: stdio
- Category: Search & AI memory
Install
Pick the AI app you use, then copy the settings.
Before you install
- Basic checks: 1 of 1 passed
- Runs on your computer with your user’s permissions (files, network, apps).
- You fill in yourself:
DARWIN_MEMO_PATH
You need
- PyPIPython’s uv installed (for the uvx command)
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json
{
"mcpServers": {
"darwin-memo": {
"command": "uvx",
"args": [
"darwin-memo",
"mcp"
],
"env": {
"DARWIN_MEMO_PATH": "your-value"
}
}
}
}Run this in your terminal
claude mcp add darwin-memo --env DARWIN_MEMO_PATH=your-value -- uvx darwin-memo mcpFill in the empty values before running it.
Or copy the settings yourself
Paste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"darwin-memo": {
"command": "uvx",
"args": [
"darwin-memo",
"mcp"
],
"env": {
"DARWIN_MEMO_PATH": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Or copy the settings yourself
Paste into .vscode/mcp.json in your project
{
"servers": {
"darwin-memo": {
"type": "stdio",
"command": "uvx",
"args": [
"darwin-memo",
"mcp"
],
"env": {
"DARWIN_MEMO_PATH": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.
Your choice stays on this device in a cookie; no account needed.
Values you fill in
DARWIN_MEMO_PATHenvPath to the persistent memory file (default ~/.darwin-memo/memory.json)
These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
What it can do
Runs on your computer; the tool list appears in your AI app after you install it.
Basic checksi
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 7, 2026
- ✓ Pass
Maker matches the repo
The registry name was verified with the GitHub account rogermsc, which owns the repo
- … Not checked yet
Still maintained
We have not read this repo from GitHub yet
- … Not checked yet
Has a licence
We have not read this repo from GitHub yet
- – Not applicable
Endpoint responds
This one runs on your computer; there is no internet endpoint
- – Not applicable
Sign-in with OAuth
Runs on your computer and uses local settings instead
- … Not checked yet
No hidden instructions in tool descriptions
Runs on your computer; we can only read tool lists from internet endpoints
Compared with others in Search & AI memoryi
- More stars than 0% of the other 7250 servers in this category
- Published to the registry more recently than 37% of the other 7250 servers in this category
- Basic checks equal to or better than 100% of the other 7250 servers in this category
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
Common questions
Is darwin-memo safe?
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
What keys or codes does it need?
The registry does not list a required key, though some servers ask you to sign in on first use.
Does it run on my computer or in the cloud?
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
Is it still maintained?
The GitHub repo last changed –, and the latest version (0.7.1) reached the registry 2 months ago.
Which apps can use it?
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Python (uv) on your computer.
What can I use instead?
In the same category: Context7, Codebase Memory, Exa
Alternatives
Servers in the same category from other makers, sorted by stars.
Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
Codebase Memory
by DeusData
Codebase knowledge graph for AI agents — 162 languages, sub-ms queries, 99% fewer tokens.
Runs on your computer★ 48,148MITUpdated 20 days ago
Install
Exa
by exa-labs
Connect AI agents to Exa for web search, content fetching, and multi-step research.
Runs in the cloud★ 5,213MITUpdated last month
Install
README from the maker
Open the README
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
darwin-memo
Measured memory for coding agents. Keep repository lessons, connect them to reported task outcomes, and inspect why each lesson stays or is removed.
Start with Python projects using pytest, GitHub Actions, and an MCP client. The core uses only the Python standard library. MCP support is an optional extra.
The product goal is lower total operating cost while preserving task success. That benefit is not established by the committed experiments. A smaller memory store or a larger energy balance is not a savings measurement.
See the mechanism in 60 seconds
From this reviewed checkout:
python -m pip install -e .darwin-memo demo --out /tmp/darwin-demo.json darwin-memo ui /tmp/darwin-demo.json
The demo uses a filesystem-backed simulation: file deletion uses file size, and protected-file recreation incurs a modeled penalty of three times that size. It does not measure restoration scratch space, I/O cost, or money. Open a lesson in the dashboard to inspect its settlements and removal reason.
To try the coding workflow with a real failing evaluation, use the fixed pytest example. Its authentication-retry defect changes one fixed case from failure to success. The example demonstrates settlement; it does not establish that memory caused the fix.
Install the development workflow
The init and task commands in this checkout are unreleased. Install them from the reviewed checkout on Python 3.10 or later:
python -m pip install -e '.[mcp]'Build the dashboard from this checkout before packaging it:
npm ci --prefix uinpm run build --prefix ui
Persistence requires POSIX advisory locks on a local filesystem. Windows and shared network stores are unsupported; the package refuses lockless persistence.
Connect one repository
- Run
darwin-memo init --profile github-pytestat your repository root. - Configure your MCP client using
.darwin-memo/mcp.example.json. - Add a repository lesson and query memory before a task.
- Bind the returned ticket to the repository, task, base commit, and fixed evaluation.
- Complete the task and commit the bound store with your changes.
- Run the generated GitHub Actions workflow against that task commit.
- Download its outcome artifact and carry the accepted store forward.
- Open the dashboard to inspect the reported outcome and retained lesson.
The GitHub pytest guide includes setup requirements and the generated operating instructions. init refuses overwrites and changes no global agent settings. The recommended MCP configuration exposes query, add, and inspection while reserving settlement and retention mutations for the host. It reduces accidental self-awarded credit; it does not protect against a malicious actor with write access to the store, evaluation, or workflow.
The workflow compares the same base evaluation at both commits. Added passing tests earn no credit. Green-to-green results earn zero. Missing observations or infrastructure failures abstain and leave the ticket pending. Bound settlements record repository, task, commit, run, evaluation fingerprint, and report hashes. Process one task at a time: independent Git snapshots do not merge.
Understand retention
A query retrieves lessons and opens a ticket. A later outcome moves the consulted lessons' bounded energy balances. Ticks charge upkeep; entries that reach the floor are removed, and similar entries can merge. Pinning exempts a lesson from removal and consolidation. The glossary explains the terms.
Settlement associates outcomes with consulted lessons. It does not establish causation. Passing tests are observable outcomes, not physically conserved resources. ci, operator, and agent identify reporting paths. Historical measured labels are caller claims; absent provenance is unknown. None of these labels authenticates a report by itself.
CLI, dashboard, and MCP mutations lock the full load–modify–save operation. MCP reloads on each call, cached embeddings survive unrelated writes, and stale Python writers are rejected. Retry the whole operation after contention. See the store format and API reference.
Inspect the evidence
| Question | Evidence | What it supports |
|---|---|---|
| What happens under corrupted feedback? | Central attack-table reconstruction, committed observations | A controlled simulated comparison with a modeled restoration penalty, including failure boundaries |
| Does a tuned counter compete? | Counter sweep, paper | Negative findings for universal superiority; tuned forgiveness can retain useful memory under small attack budgets |
| Does memory reduce total coding cost while preserving success? | Prospective protocol | An unanswered question; historical results lack the required complete cost comparison and noninferiority evidence |
| Can users complete and retain the workflow? | Pilot and release record | Pending external observation; no fabricated adoption counts |
Reconstruct the paper's central table without an installation or network:
python tools/reconstruct_paper.pyThe paper, *Attacking the Curator*, focuses on the feedback-corruption threat model, curator comparisons, and a reusable reproduction artifact. Historical replays remain replays. Read the claim audit, reproduction paths, and submission status.
Limitations and participation
Memory can be harmful, useful lessons can starve, and an attacker who controls enough outcome reports can control retention. A fixed evaluation can still be incomplete or gameable. The threat model defines the boundary. If no memory performs best in the cost study, the release will report that result.
Other integrations remain available; the GitHub pytest profile is the supported first-release path. Report setup failures with the diagnostic issue template, or choose a bounded contributor task. Release promotion, participant contact, paid experiments, and submission remain separate authorized actions.