recallLABSCO SUMMARY
A memory server you can point at an air-gapped laptop.
The distinguishing choice is that recall works with zero network calls — the default embedder ships with the package, so there is no key to get and no per-token bill for remembering things. The repository marks itself alpha and warns of breaking changes before 1.0, so pin the version you install.
What it is
A memory server your MCP client launches locally. Every conversation writes to an append-only store on disk, and the next session can search it semantically. Embeddings run offline from a model bundled with Chroma, so nothing leaves the machine unless you deliberately switch to a cloud embedder.
What you get
- Semantic search over everything stored —
recall, plusrecall_filteredfor a narrowed query andanswerfor text returned with the chunks it came from - Four ways to write, each with its own shape:
rememberfor a free-form observation,reflectfor a hypothesis-to-result artifact,anti_patternfor a looks-right-but-isn't warning,checkpointfor working state - Session continuity —
pulsereads back the latest checkpoint with reasoning and anti-patterns,session_closewrites the end-of-session reflection - Local RAG over your repo —
index_filefor one file,reindexto re-scan,snapshot_indexto persist the live store,memory_statsfor counts and sources forgetarchives instead of deleting, so removed memory is recoverable;maintenanceandbackfill_epochhandle the housekeeping- The MCP guide lists 16 tools, served identically over stdio and over SSE
Requirements
No account and no key. pip install "ai-recallworks[mcp]" (Python 3.11+), then point your client at the recall-mcp command — that is the whole config. Set RECALL_STORE_DIR if you want the store somewhere other than the default. For a shared network instance instead, run ghcr.io/recallworks/recall:latest on port 8787 with an API_KEY and talk to it over /sse or POST /tool/{name}. The package is ai-recallworks at 0.5.2; src/recall/ is MIT, the enterprise/ tree is BSL 1.1.
Setup effort
One command — pip install "ai-recallworks[mcp]"
USE CASES
01
Not re-explaining the codebase every morning
Store the conventions and decisions once with remember, then let the assistant recall them at the start of the next session instead of asking you again.
02
Answering a question with the source attached
Index the repo with index_file and reindex, then use answer so the reply arrives with the chunks it was drawn from rather than a confident guess.
03
Recording a dead end so nobody walks it twice
Write it with anti_pattern when the obvious approach turned out wrong, and it surfaces in pulse the next time work resumes.
04
Recovering memory you deleted by mistake
forget archives rather than removes, and the append-only artifacts on disk let reindex rebuild the vector store from scratch.
16 tools
Listed in project docs. Your client may show a different list.
recall_filteredanswerrememberreindexindex_filememory_statsforgetreflectanti_patternsession_closecheckpointpulsemaintenancesnapshot_indexbackfill_epochBEFORE YOU INSTALL
What to know first
Tools you get
16 tools named in the project's documentation.
Requires
recall-mcp on PATH
INSTALLATION
Connect your client
Server setup
TERMINAL
pip install "ai-recallworks[mcp]"MCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"recall": {
"command": "recall-mcp",
"env": {
"RECALL_STORE_DIR": "/Users/<YOUR_USERNAME>/.recall/store",
"RECALL_LOG_LEVEL": "WARNING"
}
}
}
}VS Code
VS CODE · JSON
{
"servers": {
"recall": {
"command": "recall-mcp"
}
}
}Cursor
CURSOR · JSON
{
"mcpServers": {
"recall": {
"command": "recall-mcp",
"env": {
"RECALL_STORE_DIR": "/Users/<YOUR_USERNAME>/.recall/store"
}
}
}
}Cline
CLINE · JSON
{
"mcpServers": {
"recall": {
"command": "recall-mcp",
"args": [],
"env": {
"RECALL_STORE_DIR": "/Users/<YOUR_USERNAME>/.recall/store"
}
}
}
}Continue
CONTINUE · JSON
{
"mcpServers": [
{
"name": "recall",
"command": "recall-mcp",
"env": {
"RECALL_STORE_DIR": "/Users/<YOUR_USERNAME>/.recall/store"
}
}
]
}Claude Desktop — Windows, Docker
Start the Recall server container on port 8787:
CLAUDE DESKTOP — WINDOWS, DOCKER · TERMINAL
docker run -d --name recall -p 8787:8787 -e API_KEY=test-key-123 -v recall-data:/data ghcr.io/recallworks/recall:latestInstall the recall-mcp bridge with the pip command above, then add this to %APPDATA%\Claude\claude_desktop_config.json:
CLAUDE DESKTOP — WINDOWS, DOCKER · JSON
{
"mcpServers": {
"recall": {
"command": "recall-mcp",
"env": {
"RECALL_BASE_URL": "http://localhost:8787",
"RECALL_API_KEY": "test-key-123"
}
}
}
}test-key-123 is the quickstart's example: API_KEY is a secret you choose when you start the container, and RECALL_API_KEY must be the same value. Here recall-mcp is a stdio bridge to the container at RECALL_BASE_URL.
