Atlas / MCP servers / intina47 / Context Sync

Context SyncBLOCK

Name: Context Sync
Author: intina47

mcp/intina47/context-sync

MCP server by intina47 · ai · ai memory · clawdbot · Audited 2026-10-06

Local persistent memory store for LLM applications including continue.dev, cursor, claude desktop, github copilot, codex, antigravity, etc.

Verdict

BLOCK

Grade

Trust score

54 /100

Exposed tools

8 5r · 3w · 0d

Transport

stdio

License

MIT

Stars

190

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

Local-first project memory for AI coding tools over MCP.

Context Sync gives an agent a persistent memory layer for a codebase:

project identity and tech stack
active work, decisions, constraints, and caveats
structured file exploration
git-aware context
optional read-only Notion lookup

What It Does

Context Sync is built for session continuity. Instead of relying on a model to remember earlier conversations, it stores the project context that matters and makes it retrievable through a small MCP tool surface.

The current core tools are:

set_project
remember
recall
read_file
search
structure
git
notion

Install

Install globally:

npm install -g @context-sync/server

Auto-configuration runs during global install. After install, restart your AI tool.

If you install locally instead of globally, auto-config does not run. Use the manual config guide in docs/CONFIG.md.

Verify

Check that context-sync appears in your MCP tool list:

Claude Desktop: open the tools list
Cursor: open the tools list
VS Code / Copilot Chat Agent mode: check tools
Continue.dev: open the Continue panel
Codex CLI: codex mcp list
Claude Code: claude mcp list

If auto-config fails, use docs/TROUBLESHOOTING.md and docs/CONFIG.md.

Optional Notion Setup

Run the setup wizard if you want Notion integration:

context-sync-setup

or:

npx context-sync-setup

Notion support is intentionally read-only at the tool surface:

notion({ action: "search", query: "..." })
notion({ action: "read", pageId: "..." })

First-Time Workflow

Start every new session by setting the project first:

1. set_project({ path: "/absolute/path/to/project" })
2. recall()
3. structure({ depth: 2 })
4. read_file({ path: "src/index.ts" })
5. remember({ type: "decision", content: "Use SQLite for local s

Read from source at commit 7e3acaa519a1OBSERVED · 2026-10-06

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.

claude-code

claude mcp add server -- npx -y @context-sync/[email protected]

claude-desktop

{
  "mcpServers": {
    "server": {
      "command": "npx",
      "args": [
        "-y",
        "@context-sync/[email protected]"
      ]
    }
  }
}

Exposed tools (8)

5 read · 3 write · 0 destructive.

Tool	Risk	Description
`context-sync-usage`	write	Complete guide on how to use Context Sync effectively as an AI agent
`debugging-context-sync`	write	How to debug Context Sync when things go wrong
`notion`	read	Access your Notion workspace documentation. Use notion.search to find pages, notion.read to view content. Essential for pulling in external documentation context.
`read_file`	read	Read a file from the current workspace. Use this to understand code context.
`recall`	read	Get context about the current project. Returns project identity, active work, constraints, problems, and recent decisions. Call this at the start of a conversation to understand what the user is working on. Essential for
`search`	read	Search the workspace. Can search by filename pattern or file contents. Use this to discover relevant files or find specific code.
`set_project`	write	Initialize a project for context tracking. Detects project type, tech stack, and architecture. Call this once when starting work on a new project. This is the foundation - run this first.
`structure`	read	Get the file/folder structure of current workspace. Use this to understand project layout.

Trust audit

BLOCKgrade F · trust 54/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

Layer	What it checks	Result
L0	Provenance & inventory	PASS
L1	Static analysis of the code	FAIL
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	PASS
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (6 observation(s))
Network: declared (1 observation(s))
Shell: declared (8 observation(s))
Dependencies: not all pinned
Secrets in source: none-found

Findings (17)

HIGHCode injection · code.deserialize · CWE-78, CWE-94, CWE-95

bin/auto-configurator.cjs:395

configObj = yaml.load(raw) || {};

Why it matters. deserialises untrusted bytes into live objects

Fix. use json or yaml.safe_load

HIGHCode injection · code.deserialize · CWE-78, CWE-94, CWE-95

bin/register-continue.cjs:61

configObj = yaml.load(raw) || {};

Why it matters. deserialises untrusted bytes into live objects

Fix. use json or yaml.safe_load

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95

bin/setup.cjs:222

exec(command);

Why it matters. evaluates text as code

Fix. remove; use a parser or a dispatch table

MEDIUMFilesystem / path · fs.system_paths · CWE-22, CWE-59

bin/platform-configs.cjs:36

'/usr/local/bin/claude',

MEDIUMFilesystem / path · fs.system_paths · CWE-22, CWE-59

bin/platform-configs.cjs:74

'/usr/local/bin/cursor',

MEDIUMFilesystem / path · fs.system_paths · CWE-22, CWE-59

bin/platform-configs.cjs:116

'/usr/local/bin/code',

MEDIUMFilesystem / path · fs.system_paths · CWE-22, CWE-59

bin/platform-configs.cjs:212

'/usr/local/bin/zed',

MEDIUMFilesystem / path · fs.system_paths · CWE-22, CWE-59

bin/platform-configs.cjs:252

'/usr/local/bin/windsurf'