Provides AI agents with repository-aware memory and context retrieval through semantic search and codebase ingestion.
Also known as cuemap-dev/cuemap-mcp
- Stars
- 1
- Tools
- 21
- Downloads, 30 days
- —
- Latest version
- 0.7.5
- Last commit
- Oct 16, 2026
- License
- MIT
What is CueMap MCP?
CueMap provides AI agents with repository-aware memory and context retrieval through semantic search and codebase ingestion. It exposes 21 tools, such as cuemap_init_preview, cuemap_init, cuemap_status.
Summary generated from the project's README
About CueMap MCP server
CueMap MCP Server provides AI coding assistants with repository-aware memory and semantic search capabilities. It allows tools like Claude Desktop, Cursor, Windsurf, and Antigravity to recall codebase context through a high-performance Rust engine. The server manages its own background processes and includes bundled models like qint8 MiniLM-L3, meaning no additional model downloads are required at runtime.
Developers can deploy this server locally with a zero-config setup using npm. It supports various operating systems including Linux, macOS, and Windows. Users can customize the environment using variables such as CUEMAP_PORT for the engine binding or CUEMAP_URL to connect to a remote server. The system handles repository ingestion through a workflow of previewing, confirming, and watching files for changes.
Beyond simple search, the server offers advanced memory management. It includes tools for project loading, saving, and unloading to optimize RAM usage. It also supports explicit ingestion of URLs, local files, or raw text, and provides specialized tools for managing aliases and lexicon relationships within the codebase memory.
What you can do with it
- 1Scan repository metadataAn AI agent can preview the files and folders in a repository before starting a full ingestion process.
- 2Perform semantic codebase searchA developer can ask an AI to find specific information within a codebase using natural language queries.
- 3Manage project memory usageAn agent can unload specific projects from memory to reduce the system's RAM consumption.
- 4Store custom natural language memoriesA user can manually save a specific piece of information or a note into the project memory.
Written from the project's README. Every use case is checked against a quote from it.
How to install CueMap MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall CueMap MCP in Claude Code
$ claude mcp add cuemap-mcp -- npx -y cuemap-mcpInstall CueMap MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"cuemap-mcp": {
"command": "npx",
"args": [
"-y",
"cuemap-mcp"
]
}
}
}Install CueMap MCP in VS Code
.vscode/mcp.json
{
"servers": {
"cuemap-mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"cuemap-mcp"
]
}
}
}Install CueMap MCP in VS Code CLI
$ code --add-mcp '{"name":"cuemap-mcp","type":"stdio","command":"npx","args":["-y","cuemap-mcp"]}'Install CueMap MCP in Claude Desktop
claude_desktop_config.json
{
"mcpServers": {
"cuemap-mcp": {
"command": "npx",
"args": [
"-y",
"cuemap-mcp"
]
}
}
}Install CueMap MCP in Codex
$ codex mcp add cuemap-mcp -- npx -y cuemap-mcpInstall CueMap MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.cuemap-mcp] command = "npx" args = ["-y", "cuemap-mcp"]
Install CueMap MCP in Gemini CLI
$ gemini mcp add -s user cuemap-mcp npx -- -y cuemap-mcpInstall CueMap MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"cuemap-mcp": {
"command": "npx",
"args": [
"-y",
"cuemap-mcp"
]
}
}
}CueMap MCP tools 21
Detected from the READMEcuemap_init_previewcuemap_initcuemap_statuscuemap_projectscuemap_project_loadcuemap_project_savecuemap_project_unloadcuemap_project_synccuemap_statscuemap_memory_getcuemap_memory_reinforcecuemap_memory_delete
| cuemap_init_preview | — | |
| cuemap_init | — | |
| cuemap_status | — | |
| cuemap_projects | — | |
| cuemap_project_load | — | |
| cuemap_project_save | — | |
| cuemap_project_unload | — | |
| cuemap_project_sync | — | |
| cuemap_stats | — | |
| cuemap_memory_get | — | |
| cuemap_memory_reinforce | — | |
| cuemap_memory_delete | — |
Show all 21 toolsShow fewer tools
cuemap_project_exportcuemap_project_artifactscuemap_ingest_urlcuemap_ingest_contentcuemap_ingest_filecuemap_addcuemap_intent_classifycuemap_recallquery_embedding
| cuemap_project_export | — | |
| cuemap_project_artifacts | — | |
| cuemap_ingest_url | — | |
| cuemap_ingest_content | — | |
| cuemap_ingest_file | — | |
| cuemap_add | — | |
| cuemap_intent_classify | — | |
| cuemap_recall | — | |
| query_embedding | — |
Configuration
CUEMAP_PORT
| Name | Description | |
|---|---|---|
| CUEMAP_PORT | — |
Automated checks
Ran Nov 13, 2026Full report →Passed our automated checks. That is not a security audit: read the code and give the server the narrowest access it needs. How checks work
- Source repository linked and reachablePass
- Open source license (OSI)Pass
- Commit or release in the last yearPass
- Package name not close to a popular packagePass
- Remote endpoint uses HTTPSNot checked
- Destructive tools are labelledPass
- No hidden instructions to the modelPass
- No critical advisory in the package or its dependenciesPass
README
View on GitHubThe Model Context Protocol (MCP) server for CueMap, allowing AI coding assistants (like Claude Desktop, Cursor, Windsurf, and Antigravity) to instantly recall codebase context using the CueMap engine.
For agent-facing operating guidance—including repository consent, supported content types, recall modes, accuracy-oriented knob selection, and troubleshooting—see SKILL.md.
Zero-Config Deployment
The CueMap MCP Server is designed to work completely out-of-the-box. When started, it automatically manages a high-performance Rust instance of the CueMap Server in the background. The v0.7.3 engine bundles qint8 MiniLM-L3 by default and q4 MiniLM-L3 for the edge profile; no model download occurs at runtime.
You do not need to install or run the CueMap CLI manually. The correct pre-compiled binary for your operating system is automatically downloaded via optional NPM dependencies. Embedded startup supports Linux x64/ARM64, macOS x64/ARM64, and Windows x64.
Installation
# Global installation makes the `cuemap-mcp` command available
npm install -g cuemap-mcp
(Note: Ensure your package manager is configured to download optionalDependencies so the local Rust binary is included).
Agent Plugin
CueMap also ships a separate cuemap-agent-plugin package for Agent Plugins-compatible clients. It bundles the portable plugin.json manifest, stdio mcp.json configuration, and a repository-memory skill. The plugin launches this MCP server at the matching release version and must be installed or downloaded separately from cuemap-mcp.
Configuration (Environment Variables)
By default, the embedded engine runs on port 8735. You can customize the server behavior by passing the following environment variables in your MCP configuration:
CUEMAP_PORT: Override the port the embedded engine binds to (default:8735).CUEMAP_CONFIG_PATH: Absolute path to a customserver_config.tomlto configure advanced Engine tuning, background jobs, and RAG search parameters.CUEMAP_URL: If you prefer to bypass the embedded engine and connect to a remotely hosted or separately running CueMap server, specify its URL here (e.g.http://localhost:8735).CUEMAP_PROJECT: Override the default project. Without this setting, CueMap derives a stable repository-scoped project from the current Git remote or working directory.CUEMAP_LOG_PATH: Override the embedded engine log path. It defaults to~/.cuemap/server.log, which is the file read bycuemap logs.
Server lifecycle
The MCP server calls EmbeddedCueMap.start() as soon as Codex launches the plugin process, before any CueMap tool is invoked. It attaches to a compatible engine already listening at CUEMAP_URL or the configured port; otherwise it starts and owns a local engine automatically. cuemap_init does not start the server—it previews or applies a confirmed repository scope after the engine is ready. An engine owned by the MCP process is stopped gracefully when that process exits.
Using with AI Agents
To use this MCP server with your AI assistant, add it to your assistant's MCP configuration file.
Example Configuration (Claude Desktop)
{
"mcpServers": {
"cuemap": {
"command": "npx",
"args": [
"-y",
"cuemap-mcp"
],
"env": {
"CUEMAP_PORT": "8735"
}
}
}
}
Available Tools
-
cuemap_init_preview: Performs a read-only repository metadata scan and returns supported top-level folders/root files, counts, the stable repository project ID, and any existing saved scope. Present this preview to the user before ingestion.path(string): Absolute path to the repository root.projectName(string, optional): Project ID override. Defaults to the stable repository-scoped ID.includedPaths(string[], optional): Repository-relative files or folders to preview. Empty means all supported files allowed by ignore rules.ignoredPatterns(string[], optional): Additional gitignore-style exclusions.ignoredExtensions(string[], optional): Additional excluded extensions.
-
cuemap_init: Applies a user-confirmed repository ingestion scope, starts the initial scan, and keeps watching that scope for new, changed, deleted, newly ignored, or newly unignored supported files. It only reports verified completion after observing ingestion activity, reaching a terminal job phase, and completing intent annotation; an initialidlestatus with0/0writes is not considered complete.path(string): Absolute path to the repository root.projectName(string, optional): Project ID override. Defaults to the stable repository-scoped ID.includedPaths(string[], optional): User-approved repository-relative files or folders. Empty means all supported files allowed by ignore rules.ignoredPatterns(string[], optional): Additional gitignore-style exclusions.ignoredExtensions(string[], optional): Additional excluded extensions.confirmed(boolean): Must be true only after explicit user confirmation of the previewed scope.
-
cuemap_status: Returns project-scoped background ingestion progress plus derived monitoring fields. Poll it aftercuemap_inituntilverified_completeistrue.project(string, optional): Project ID override. Defaults to the stable repository-scoped ID.active: Whether the current phase is non-terminal or writes/intent annotations remain pending.observed_activity: Whether this MCP process has observed an active phase or positive write/intent counts for the project.pending_writes: Remaining writes based on the latest totals.pending_intents: Remaining memory intent annotations based on the latest totals.verified_complete: True only after activity has been observed, the project reachesdoneoridle, no work failed or remains pending, andintent_readyis not false.
Project inspection and memory lifecycle
cuemap_projects: List projects and their summary metadata, including whether each project is currently loaded in RAM.cuemap_project_load: Explicitly warm a persisted project before a latency-sensitive operation. Ordinary project requests demand-load it when needed.cuemap_project_save: Persist a current snapshot without unloading it.cuemap_project_unload: Persist and unload a project to reduce memory usage. Active work can produce a retryable busy response.cuemap_project_pack/cuemap_project_package_load: Write or load a ready-to-query local.cuemappackage.cuemap_project_push/cuemap_project_pull: Transfer a package through S3 using the engine host's configured AWS CLI.cuemap_project_sync: Fast-forward immutable S3 history and refuse divergence.cuemap_stats: Read repository-project statistics, or global engine statistics withglobal: true.cuemap_memory_get: Read one memory as decoded text and provenance by numericmemory_idand owningproject. Requires an engine supportingGET /memories/:id?decoded=true; older engines return an upgrade error through this tool.cuemap_memory_reinforce: Reinforce one memory, optionally on explicitcues.cuemap_memory_delete: Permanently delete one memory. Requiresconfirmed: trueafter explicit user confirmation.cuemap_project_export: Export a cursor-paginated project page with configurable content, cue, and metadata inclusion.cuemap_project_artifacts: Inspect CueBridge artifact metadata without reloading it.
Explicit ingestion
cuemap_ingest_url: Ingest one URL or recursively crawl it withdepthandsame_domain_onlycontrols.cuemap_ingest_content: Ingest supplied text with filename, source-key, metadata, structural-cue, segmentation, and optional one-vector-per-produced-chunkembeddings.cuemap_ingest_file: Ingest one user-approved local file path.
These tools are for explicit ingestion requests. Repository initialization continues to use the preview, confirmation, and watcher workflow above.
Alias and Lexicon administration
-
cuemap_alias_list,cuemap_alias_add,cuemap_alias_merge: Inspect and manage manual cue aliases. Merging requiresconfirmed: true. -
cuemap_lexicon_inspect,cuemap_lexicon_graph,cuemap_lexicon_wire,cuemap_lexicon_delete: Inspect and administer Lexicon relationships. Deletion requiresconfirmed: true. -
cuemap_add: Stores one natural-language memory. The project is created automatically if it does not exist.content(string): Memory content to store.project(string, optional): Project ID that will own the memory. Defaults to the current repository-scoped project.cues(string[], optional): Explicit cues/tags to associate with the memory.metadata(object, optional): JSON metadata to store with the memory.source_key(string, optional): Stable source key for deterministic upsert/deduplication.event_time(number, optional): Original event timestamp as Unix seconds. Defaults to ingestion time.embedding(number[], optional): Precomputed memory embedding.disable_temporal_chunking(boolean, optional): Disable temporal chunking for this memory.async_ingest(boolean, optional): Process ingestion in the background and return immediately.
-
cuemap_intent_classify: Classifies text with the engine's local intent model. Acceptstarget: "query"or"memory"and returns relative intent scores, recall/memory eligibility, confidence weight, and model/taxonomy versions. Scores are ranking signals, not calibrated probabilities. -
cuemap_recall: Recalls context about a codebase from your CueMap integrated brain. Uses natural language and semantic search to find relevant information.query(string): The natural language query to search for.limit(number, optional): Maximum results to return (default: 10).response_mode(full|preview, optional): Defaultfull. Usepreviewfor broad discovery; each hit returns a leading excerpt instead of full content, retaining IDs and metadata.preview_chars(integer, optional): Preview length cap, 100–2000 UTF-16 code units (default: 200). Ignored in full mode. Metadata and diagnostics are not capped.projects(string[], optional): List of project IDs to scope the search to. Multiple enables cross-project queries.cues(string[], optional): Specific cue tags to filter the search.query_time(string, optional): Timestamp or natural-language time anchor for v0.7 temporal query intent.depth(number, optional): Depth of multi-hop recall expander (default: 1).expansion_depth(number or"conversation_round", optional): Neighbor context expansion (default:"conversation_round"). Values above 1 include nearby parent chunks or source-ordered context, using a radius ofexpansion_depth - 1when linkage exists."conversation_round"prepends the message each result replies to (for an assistant reply, the user message that prompted it) when memories carry session, turn and role metadata, for example after ingesting withsegmenter: "conversation". The top result also gets the message that replies to it (for a user message, the assistant's answer). Memories without conversation metadata are returned as they are; pass1for matched memories only. A message that would also appear lower in the results is moved up instead of repeated, so fewer thanlimitresults can come back.auto_reinforce(boolean, optional): Automatically reinforce retrieved memories (default: false).min_intersection(number, optional): Minimum required cue intersection count (default: 0).explain(boolean, optional): Include scoring explanation data in results (default: false).trace_timing(boolean, optional): Include v0.7 timing diagnostics.disable_salience_bias(boolean, optional): Disable salience scoring bias.disable_alias_expansion(boolean, optional): Disable lexicon synonym injection during querying (default: true).parent_fusion(off|auto|force, optional): Chunk-parent fusion mode.ordered_reconstruction(off|auto|force, optional): Ordered session reconstruction mode.evidence_coverage(off|auto|force, optional): Multi-evidence coverage mode.disable_cuebridge_artifacts(boolean, optional): Disable CueBridge artifact expansion.cuebridge_gap_limit(number, optional): Maximum CueBridge gap expansions.semantic_mode(lexical,semantic, orhybrid, optional): Choose cue-only recall, vector candidate discovery, or local semantic reranking of lexical candidates. The engine default ishybrid.
-
query_embedding(number[], optional): Supply a precomputed query vector when the calling application owns the embedding provider.
Recall response and follow-up inspection
Recall returns engine JSON in both MCP structuredContent and a matching JSON
text block. This replaces the earlier prose-only format. Each identified
memory carries project_id and memory_id; metadata, source locations, and
requested diagnostics are preserved as supplied by the engine. Explicit
project queries may return project blocks containing their own results or
error, including empty groups. Ordinary unscoped queries return memory
records directly in results.
Use a hit's project and ID with cuemap_memory_get when the stored record is
needed. That tool returns readable content without vectors or compressed
storage bytes; it does not expand neighboring chunks or read live files.
Prefer opening the cited source when recall already returned the relevant
text. Follow-up recalls should address a new evidence gap. See SKILL.md
for the coding investigation workflow and bounded recall controls.
For broad discovery, pass response_mode: "preview". A hit's content is
replaced by preview, content_truncated, and content_length (UTF-16 code
units). This is a leading excerpt, not a generated summary or query-selected
snippet. Full content is omitted from both MCP output forms; fetch promising
memories by their project/ID or read the live source. Full mode remains the
default. The engine shapes previews before sending its response. This does not change
ranking or retrieval work, and does not limit metadata or diagnostics.
License
The MCP server package is MIT-licensed. Its optional native engine dependencies are separate packages: v0.7.3 and later engine packages are Apache-2.0, while pre-v0.7.3 engine packages remain under BSL-1.1. See LICENSE for the MCP server license.
CueMap MCP questions
How do I add CueMap MCP to Claude Code?
Run claude mcp add cuemap-mcp -- npx -y cuemap-mcp in your terminal. Add --scope user to make it available in every project.
How do I add CueMap MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
What tools does CueMap MCP provide?
It exposes 21 tools, including cuemap_init_preview, cuemap_init, cuemap_status, cuemap_projects, cuemap_project_load.
Do I need to install the CueMap CLI manually?
No, the correct pre-compiled binary for your operating system is automatically downloaded via optional NPM dependencies.
Does the server require a model download at runtime?
No, the engine bundles MiniLM-L3 models by default, so no model download occurs during runtime.