Searching Sourcegraph

Searching Sourcegraph

You have access to Sourcegraph search and code navigation tools through MCP.

Tools

• nls_search({ query }): semantic search for natural-language exploration.
• keyword_search({ query }): keyword or regex search using Sourcegraph query syntax.
• read_file({ repo, path, startLine?, endLine?, revision? }): fetch file contents.
• list_repos({ query, limit?, after?, before? }): discover relevant repositories.
• list_files({ repo, path?, revision? }): list files and directories.
• go_to_definition({ repo, path, symbol, revision? }): locate symbol definitions.
• find_references({ repo, path, symbol, revision? }): find symbol usage.
• commit_search({ repos, messageTerms?, authors?, contentTerms?, files?, after?, before? }): search commit history.
• diff_search({ pattern, repos, added?, removed?, author?, after?, before? }): search code diffs.
• compare_revisions({ repo, base, head, first?, after? }): compare two revisions.
• get_contributor_repos({ author, limit?, minCommits? }): discover contributor activity by repo.
• deepsearch({ question }): run Sourcegraph Deep Search for complex, multi-step questions.
• deepsearch_read({ identifier }): read a Deep Search conversation by URL or token.

Workflow

1. Start with a constrained search (repo:, file:, lang:, type: when possible).
2. If results are broad, add constraints and rerun.
3. Read promising files with read_file to confirm behavior.
4. Iterate queries based on findings.

Prefer nls_search for exploration and keyword_search for precision.

Query Tips

• Use repo:^github\\.com/org/repo$ for exact repository matching.
• Use file:\\.(go|ts|py)$ to constrain by extension.
• Use type:symbol for definitions and type:file for content matches.