About Context Vault
Context Vault gives AI agents persistent memory across sessions, storing entries as local Markdown files that people can edit. It combines full-text and semantic search, with an optional hosted tier for cloud sync, a web dashboard, and team sharing.
Key features
- 01Store memory as Markdown files
- 02Hybrid full-text and semantic search
- 03Browse by kind, tag, or date
- 04Ingest pages and local projects
- 05Compile topic entries into briefs
- 06Hosted cloud sync and team sharing
Use cases
- 01Recall project decisions and references in a new session
- 02Search saved clues from errors or file paths
- 03Register web pages or project contents as memory
- 04Summarize multiple entries around a topic
MCP support
Context Vault provides MCP tools for agents to save, search, browse, and delete memories, ingest web pages and projects, and create topic briefs. Its setup command registers the server in the MCP configurations of detected AI tools.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://api.context-vault.com/mcpclaude mcp add --transport http fellanh-context-vault https://api.context-vault.com/mcp{"mcpServers":{"fellanh-context-vault":{"type":"http","url":"https://api.context-vault.com/mcp"}}}Open in Cursor{"servers":{"fellanh-context-vault":{"type":"http","url":"https://api.context-vault.com/mcp"}}}Open in VS Code{"mcpServers":{"fellanh-context-vault":{"type":"http","url":"https://api.context-vault.com/mcp"}}}{"mcpServers":{"fellanh-context-vault":{"serverUrl":"https://api.context-vault.com/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEcontext-vault
Persistent memory for AI agents: your agent saves what it learns and finds it again in the next session. Entries are plain markdown files on your machine.
Hosted tier: cloud sync, web dashboard, team sharing at app.context-vault.com
Quick start: a memory recalled in a new session, in under 5 minutes
Timed on a clean macOS account with Claude Code: about one minute of commands, plus your typing.
1. Install the binary (macOS and Linux; no Node needed):
curl -fsSL https://context-vault.com/install.sh | bash
It puts context-vault in ~/.local/bin. If it says that folder is not on your PATH, run the line it prints, then open a new terminal.
2. Connect your AI tools:
context-vault setup
This adds Context Vault to the MCP config of Claude Code and every other client it finds (Claude Desktop, Cursor, Windsurf, Cline, Roo Code). --client claude-code limits it to one; --dry-run shows what it would write.
3. Restart your AI tool. Quit it and start it again (for Claude Code, a new claude in a new terminal). The first start downloads the search model (about 90 MB) once.
4. Save something. Name the vault, or Claude Code may keep it in its own built-in memory instead:
Save to my vault: our staging bucket is called aurora-gravel-17
5. Recall it in a new session. Quit, start a fresh session, and ask:
Search my vault: what is our staging bucket called?
The answer comes from the vault, and the entry is a markdown file under ~/.vault/ that you can open and edit. context-vault doctor should print engine health: OK.
Windows: irm https://context-vault.com/install.ps1 | iex, open a new terminal, then steps 2 to 5.
What it does
- Save: insights, decisions, patterns, references. Your agent writes them as you work.
- Search: hybrid full-text and semantic search, in natural language.
- Own your data: markdown in a folder you control, git-friendly and human-editable.
MCP tools
Your agent calls these; you talk to it in plain words.
| Tool | What it does |
|---|---|
save_context | Save an entry, or update one by id |
get_context | Search the vault (full-text and semantic) |
recall | Light hints from a raw signal: a prompt, an error, a file path |
list_context | Browse entries by kind, tags or date |
delete_context | Remove an entry by id |
ingest_url | Fetch a page and save its readable text |
ingest_project | Scan a local project and register it |
create_snapshot | Compile entries on a topic into one brief |
session_end | Save a session summary and its discoveries |
context_status | Vault health: paths, counts, issues |
list_buckets | List buckets (a bucket:<name> tag scopes entries) |
Kinds (insight, decision, pattern, reference, or your own) are folders in the vault.
CLI
| Command | What it does |
|---|---|
context-vault setup | Register the MCP server with your AI tools |
context-vault doctor | Check the install: paths and engine health |
context-vault status | Entry counts and resolved paths |
context-vault search <q> | Search from the shell |
context-vault save | Save an entry from the shell |
context-vault reindex | Rebuild the index from the markdown files |
context-vault --help | Every command |
Configuration
| Setting | Default | Override |
|---|---|---|
| Vault | ~/.vault/ | CONTEXT_VAULT_VAULT_DIR |
| Data | ~/.context-mcp/ | CONTEXT_VAULT_DATA_DIR |
| Database | ~/.context-mcp/vault.db | CONTEXT_VAULT_DB_PATH |
Troubleshooting
context-vault: command not found: the install folder is not on your PATH. Run~/.local/bin/context-vault setup, or add the line the installer printed to your shell profile.- Your agent says the vault tools are missing: restart the AI tool after setup; in Claude Code,
/mcpshows whethercontext-vaultconnected.context-vault setup --printshows the entry it writes. - Your agent saved it somewhere else: say "my vault" in the request.
- Stale search results:
context-vault reindex.
The npm package (npx context-vault) is still the older 3.20 Node build; use the install script above.
Hosted option
Cloud sync, web dashboard and team sharing at app.context-vault.com. Setup for Claude Code, Cursor and GPT Actions: docs/distribution/connect-in-2-minutes.md
Feedback
- Questions or ideas? Discussions
- Found a bug? Open an issue
Development
git clone https://github.com/fellanH/context-vault.git
cd context-vault
cargo build --release -p context-vault-cli
./target/release/context-vault --help
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.