sqry - Semantic Code Search

io.github.verivus-ossPublished by GitHub account verivus-ossv33.0.1Updated Oct 24, 2026

AST-based semantic code search: 39 tools over 37 languages. Call tracing and impact analysis.

VerifiedSTDIODesktop onlyKnowledge & MemoryDeveloper Tools

Overview

AI-generated overview

sqry exposes AST-based semantic code search over a local repository, letting an assistant find symbols, trace calls and analyze impact across 37 languages.

What you can do
  • Search and list code symbols by kind, language and visibility using string predicates and structured filters
  • Navigate definitions and references, and run relation queries across the code graph
  • Export graphs and trace calls for impact analysis and semantic diffs
  • Detect duplicates, cycles and unused code, and check workspace status
  • Discover available tools and schemas through tools/list, a list-tools flag and sqry resource documents
When to use it
Useful when an assistant needs to understand or navigate a large multi-language codebase: locating definitions, following call chains, judging the blast radius of a change, or finding dead and duplicated code. It runs locally against a repository you point it at.
Requirements
A local install of the sqry-mcp binary (Cargo package or container image) run over stdio, plus the sqry CLI for setup and daemon modes. The SQRY_MCP_WORKSPACE_ROOT environment variable sets the repository to index and search. Desktop-only; no authentication is declared.
Before you install
The server indexes and reads your source tree, so results can expose file paths and code to the assistant and, if it uses an external model provider, to that provider; the README recommends the standard redaction preset for hosted or external LLM providers. Setup commands write MCP configuration files, so preview with the dry-run option first. Daemon mode requires starting and loading sqryd deliberately.

Installation

In SourceWeft

  1. Open sqry - Semantic Code Search in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

sqry MCP Server

Version: 33.0.1

MCP Registry name: mcp-name: io.github.verivus-oss/sqry

sqry-mcp exposes sqry semantic code-search tools to Model Context Protocol clients.

Modes

Standalone mode:

bash
sqry-mcp --no-daemon

Daemon-backed mode:

bash
sqry daemon startsqry daemon load .sqry-mcp --daemon

Standalone mode currently exposes 39 tools. Daemon-hosted MCP exposes a 17-tool subset backed by sqryd. Use dynamic discovery for exact schemas and descriptions:

bash
sqry-mcp --list-tools

MCP clients can also call tools/list, and sqry clients can read sqry://meta/manifest.

Removed in 21.0.0: the sqry_ask natural-language MCP tool was removed. Use sqry_query, semantic_search, and relation_query instead; see Removed features for migration.

Setup

bash
sqry mcp setup --tool claudesqry mcp setup --tool codexsqry mcp setup --tool gemini

Preview without writing:

bash
sqry mcp setup --tool codex --dry-run

Codex and Gemini use global MCP config and rely on starting the assistant in the target project. Claude project scope can pin a workspace root.

When setup writes a standalone command, expect --no-daemon. Use --daemon only when you have started and loaded sqryd intentionally.

Tool Discovery

Prefer discovery over copying static tables:

  • MCP tools/list
  • sqry-mcp --list-tools
  • sqry://meta/manifest
  • sqry://docs/tool-guide
  • sqry://docs/capability-map

The standalone catalog includes tools for symbol search, navigation, relation queries, graph export, impact analysis, semantic diff, duplicate/cycle/unused detection, workspace status, and context propagation.

Quick Start Tool Calls

Use query for string predicates and filters for structured JSON pre-filters:

json
{  "name": "semantic_search",  "arguments": {    "query": "kind:function items:true",    "filters": {      "language": ["rust"],      "visibility": "public"    }  }}

Use items:true or is_definition:true inside query for definition-only predicates. For list_symbols, use the structured items_only parameter instead:

json
{  "name": "list_symbols",  "arguments": {    "kind": "function",    "language": "rust",    "items_only": true  }}

On pre-V16 snapshots without trustworthy definition signal, these definition-only surfaces return a reindexRequired / reindex-required advisory instead of silently evaluating stale default-false marker data.

Workspaces

MCP resolves workspace context from explicit paths, file-bearing arguments, MCP roots, recent session context, and current working directory fallback.

For multi-root projects, configure a workspace with .sqry-workspace or a VS Code .code-workspace file containing sqry.workspace.

bash
sqry workspace status . --json

source_root_id

workspace_status.aggregate.source_root_statuses[].source_root_id is an opaque 8-hex display/correlation token. It is not a path and not a path prefix.

Invalid tool path:

text
485f1995/src/lib.rs

Valid tool paths:

text
src/lib.rs/absolute/path/to/workspace/src/lib.rs

Under redaction, result paths can be rendered with a source-root prefix for display. Use source_root_id only to correlate that display prefix with aggregate status entries. Cleartext source roots appear only in top-level source_roots[] when the selected redaction preset permits them.

Redaction

The MCP runtime default is minimal. For hosted or external LLM providers, standard is the recommended preset unless stricter path privacy is required. strict is available for stronger redaction.

Keep these separate:

  • Runtime default: minimal
  • Recommended external-provider preset: standard
  • Library/environment defaults: see sqry-mcp-redaction docs for embedding-specific behavior

Related Guides

Source: sqry-mcp/README.md at commit 8f139a6

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

4
  1. v33.0.1LatestOct 6, 2026
  2. v33.0.0Oct 5, 2026
  3. v32.0.1Oct 1, 2026
  4. v31.0.0Sep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts