mcp server

Octocode MCP - AI Context Platform

AI code research platform. Search, analyze, and extract insights from any GitHub repository.

Description as published by the maintainer. Source

• version 7.0.9
• active
• data extraction
• retrieval

active — Most recent push to the repository was 2026-08-05. Dashed tags are derived by ZBS Index from the published description, not stated by the maintainer.

What this server can do

5 functions, named and described by the server itself. Parameter names are shown because they say more about what a function does than its name usually does.

githubGetFileContent(queries)

FILE READER - Read file contents with smart extraction PARAMS: See schema for parameter details PURPOSE: Read file contents using patterns, line ranges, or full file mode USE_WHEN: Have file path | After githubSearchCode/githubViewRepoStructure discovery AVOID: Unknown location - search first WORKFLOW: Step 1: Discovery (githubSearchCode/githubViewRepoStructure) Step 2: Read (matchString or line range) Step 3: Explore related patterns STRATEGY (choose based on what you know): - matchString: Know function/class/pattern name → start here [most efficient] - startLine + endLine: Know location from prior search → precise extraction - fullContent: Need entire file → use ONLY for small files (<500KB) OPTIMIZATION: - Start lean: Single matchString query first - If related context needed: Bulk queries for multiple files (e.g., auth.ts + user.ts + session.ts in parallel) GOTCHAS: - minified=true breaks JSON/YAML formatting → use minified=false for config files - matchString returns empty if no matches → try broader pattern or use fullContent - Files >500KB: MUST use matchString or line ranges, fullContent will truncate NEXT_STEP: hasResults → githubSearchCode for related patterns empty → githubSearchCode/githubViewRepoStructure to locate file EXAMPLES: matchString="validateUser", matchStringContextLines=20 # BEST: targeted read startLine=1, endLine=100 # Line range fullContent=true # Full file queries=[{path:"a.ts",matchString:"fn1"},{path:"b.ts",matchString:"fn2"}] # Bulk path="config.json", fullContent=true, minified=false # Config file, keep formatting GUARDS: config? - minified=false | know pattern? - matchString | large file? - use line range | efficiency? - prefer startLine/endLine over fullContent, justify range in reasoning Required: queries.

githubSearchCode(queries)

CODE SEARCH - Search file content or filenames/paths PARAMS: See schema for parameter details PURPOSE: Search file contents for code patterns or search file/directory names USE_WHEN: Know patterns | Need discovery AVOID: Broad terms | No owner/repo (rate limits) WORKFLOW: Step 1: Discovery (match="path") - Find files [fastest, most token efficient] Step 2: (Optional) Use githubViewRepoStructure to understand structure when exploring unfamiliar repos Step 3: Detailed (match="file", limit=5) - Get matches with context Step 4: Read (githubGetFileContent matchString) - Full content STRATEGY: - match="path": Fast discovery, returns paths only - match="file": Detailed search, returns text_matches[] with context - Always scope: owner/repo to avoid rate limits OPTIMIZATION: - Start lean: Know exact term? → Single specific query (e.g., "authenticate") - Uncertain about terms? → Bulk query with variants upfront (e.g., [{keywordsToSearch:["auth"]}, {keywordsToSearch:["authenticate"]}, {keywordsToSearch:["credential"]}]) - Semantic variants: "auth"→"authorization"/"authenticate"/"credential", "config"→"configuration"/"settings"/"options" GOTCHAS: - match="file" without limit = token explosion - No owner/repo = rate limits - text_matches show exact locations → copy match text into githubGetFileContent matchString for precise extraction NEXT_STEP: hasResults → githubGetFileContent matchString for full context empty → githubViewRepoStructure or broaden search EXAMPLES: match="path", keywordsToSearch=["auth"] # Fast: find files with "auth" in path owner="facebook", repo="react", keywordsToSearch=["useState"] # Content search path="src/api", extension="ts", keywordsToSearch=["export", "function"] # Precise queries=[{keywordsToSearch:["termExample"],match:"path"},{keywordsToSearch:["patternExample"],match:"file",limit:5}] # Bulk: combine discovery + detailed search keywordsToSearch=["validateUser"], match="file", limit=5 # Detailed matches GUARDS: discovery? - match="path" | many results? - limit=5-10 | no scope? - add owner/repo Required: queries.

githubSearchPullRequests(queries)

PR SEARCH - Search or fetch PRs PARAMS: See schema for parameter details PURPOSE: Get PRs with metadata, discussions, and diffs USE_WHEN: PR context | Implementation research | Review merged changes AVOID: Current code → githubGetFileContent | Code patterns → githubSearchCode WORKFLOW: Step 1: Discover (prNumber or filters) Step 2: Analyze (withComments/withContent) Step 3: Compare (githubGetFileContent for current) STRATEGY: - prNumber: Direct fetch [fastest] when you know PR number - Search: Filter by state/author/labels/dates for discovery - withContent/withComments: Start without them, add only when needed - limit: Use 3 for initial scan, 5-10 for comprehensive review OPTIMIZATION: - Start lean: Know PR number? → prNumber (fastest, most direct) - Searching? → Start with limit=3, no withContent/withComments - Need details? → Sequential: analyze metadata first, then fetch with withContent=true if relevant - Multiple PRs: Bulk queries (e.g., [{prNumber:123}, {prNumber:456}, {prNumber:789}]) GOTCHAS: - merged=true needs state="closed" - prNumber ignores other filters - label accepts string or array for OR matching NEXT_STEP: hasResults → githubGetFileContent for current | githubSearchCode for patterns empty → Broaden filters or change state EXAMPLES: prNumber=123 # FASTEST state="open", limit=5 # Recent state="closed", merged=true, withContent=true # Implementation research author="username", state="closed", merged=true # Expert contributions query="auth", match=["title","body"], limit=3 # Text search queries=[{prNumber:123},{prNumber:456}] # Bulk GUARDS: implementation? - withContent=true | discussion? - withComments=true | production? - merged=true Required: queries.

githubSearchRepositories(queries)

REPOSITORY SEARCH - Search repos by keywords/topics PARAMS: See schema for parameter details PURPOSE: Discover GitHub repositories. Gateway for codebase exploration. USE_WHEN: Starting research | Finding projects/libraries AVOID: Know repo name - go to githubViewRepoStructure | Need code patterns - use githubSearchCode WORKFLOW: Step 1: Discover (topicsToSearch or keywordsToSearch) Step 2: Explore (githubViewRepoStructure) Step 3: Search (githubSearchCode) Step 4: Read (githubGetFileContent) STRATEGY: - topicsToSearch: Most precise for curated topic tags - keywordsToSearch: Flexible search across name/desc/README - Always add stars filter for quality (e.g., stars=">1000") OPTIMIZATION: - Start lean: Know exact topic? → topicsToSearch with stars filter (e.g., topicsToSearch=["typescript"], stars=">1000") - Uncertain? → Bulk with variants: topics + keywords in parallel (e.g., [{topicsToSearch:["cli"]}, {keywordsToSearch:["command-line"]}]) - Semantic variants: "auth"→"authentication"/"oauth", "ai"→"machine-learning"/"llm" - Quality: stars=">1000" for production-ready, sort="stars" for popularity GOTCHAS: - topicsToSearch: use popular topics (e.g., "mcp", "react", "nextjs", "langchain") for best results - size filter uses KB not MB ("1000" = 1MB) - No filters = generic, low-quality results NEXT_STEP: hasResults → githubViewRepoStructure or githubSearchCode empty → Try semantic variants or broaden stars filter EXAMPLES: topicsToSearch=["typescript", "cli"], stars=">1000" # BEST: curated quality repos keywordsToSearch=["authentication", "jwt"], stars=">500" # Keyword search with quality owner="facebook", sort="stars", limit=10 # Org's popular repos keywordsToSearch=["react"], match=["name"], stars=">1000" # Name-only search created=">=2024-01-01", sort="updated", limit=5 # Recent active projects queries=[{topicsToSearch:["mcp"]},{keywordsToSearch:["mcp"]}] # Bulk parallel GUARDS: discovery? - topicsToSearch+stars | quality? - stars=">1000" | maintained? - updated filter Required: queries.

githubViewRepoStructure(queries)

DIRECTORY EXPLORER - Understand codebase organization PARAMS: See schema for parameter details PURPOSE: Display directory structure with file sizes USE_WHEN: New codebase | Need structure overview AVOID: Know filename - use githubSearchCode | Need content - use githubGetFileContent WORKFLOW: Step 1: Overview (depth=1, path="") Step 2: Drill down (depth=2, path="src") Step 3: Search (githubSearchCode) Step 4: Read (githubGetFileContent) STRATEGY: - depth=1: Fast overview, see immediate files/folders - depth=2: Deeper view, includes subdirectories (slower) - path="": Start at root, then drill into specific dirs OPTIMIZATION: - Start lean: depth=1 for overview, only go depth=2 if structure unclear - Bulk queries: Explore key directories in parallel (e.g., [{path:"src"}, {path:"tests"}, {path:"docs"}]) - Use for discovery, then githubSearchCode for content GOTCHAS: - depth=2 on dirs with >50 subdirs = slow, stick to depth=1 - path format: "src/api" NOT "/src/api" (no leading slash) - Returns files[] and folders[], no content - Dirs with >100 items: result truncates at 100 → use githubSearchCode with path filter instead NEXT_STEP: hasResults → githubSearchCode or githubGetFileContent empty → Check parent or verify repo EXAMPLES: owner="facebook", repo="react", branch="main", path="", depth=1 # Root overview owner="facebook", repo="react", branch="main", path="src", depth=2 # Deep dive queries=[{owner:"facebook",repo:"react",branch:"main",path:""},{owner:"facebook",repo:"react",branch:"main",path:"src"}] # Bulk owner="facebook", repo="react", branch="main", path="packages/react-dom", depth=1 # Monorepo GUARDS: new? - depth=1, path="" | large? - depth=1 first | monorepo? - explore each package Required: queries.

Last successful function declaration observed on 2026-08-13. Source: pkg:npm/octocode-mcp. We list what the server declared; we do not call any of these functions.

Endpoint status observed on 2026-08-13. Source: pkg:npm/octocode-mcp.

Signals

These are separate measurements of different things. They are deliberately not combined into one score, because a popularity number that mixes website traffic with saves and stars cannot be checked or acted on.

Where to get it

• Repository — https://github.com/bgauryy/octocode-mcp
• npm: octocode-mcp

Related, by what their authors tagged them

• com.clauxel.speckitacceptance/speckitacceptance-mcp — last commit 2026-05-20, shares development
  Paid remote MCP for spec driven development acceptance gate MCP, structured receipts, audit logs, an
• io.github.crypto-ninja/github-mcp-server — last commit 2026-07-28, shares github-api
  AI-optimized GitHub MCP server: 112 tools, 98% token reduction, compact responses.
• ivygrep — last commit 2026-08-06, shares code-intelligence, code-search, semantic-search
  Local code search and task context packs for coding agents
• io.github.containers/kubernetes-mcp-server — last commit 2026-08-06, shares context, modelcontextprotocol
  A Model Context Protocol (MCP) server for Kubernetes and OpenShift
• CodeNib — last commit 2026-08-06, shares code-intelligence, code-search
  Ranked codebase search and static symbol navigation for coding agents.
• coldstart — last commit 2026-08-06, shares code-intelligence, code-search
  Codebase memory for AI agents: an AST index plus agent-written notes that self-stale.
• Synap Memory — last commit 2026-08-06, shares context
  Persistent memory for AI agents — log and recall conversation context over MCP.
• GrantAi Memory — last commit 2026-03-16, shares context
  Persistent memory for AI agents. Infinite context with sub-millisecond recall.
• dev.ctxfile/ctxfile — last commit 2026-08-03, shares context
  Local-first MCP server that snapshots your project's working state into one context object.
• AllocContext — last commit 2026-07-10, shares context
  Self-host MCP: portfolio crypto context for agents. PyPI stdio; MIT.

These share tags the maintainers applied themselves, such as development, github-api, code-intelligence, code-search. Common tags like "mcp" or "ai" are ignored for this: agreeing with six hundred other projects is not a similarity.

This is not a recommendation and not a test result. It is a map of what the authors said their work is about.

How the author describes it

Topics the maintainer set on GitHub: agent, ai, ai-agents, ai-tools, claude-ai, code-intelligence, code-search, context, cursor, cursor-ai, development, github, github-api, llm, mcp, model-context-protocol, modelcontextprotocol, octocode, semantic-search, vibe-coding.