# Map or search a large or unfamiliar codebase

> Map or search a large or unfamiliar codebase: 0 companies fully do it, 8 with a limit, 25 carry the tag. Every listing in Code & developer platforms with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/code.understand-codebase/
- Markdown: https://www.anchorterminal.com/niches/code.understand-codebase/index.md (~1,550 tokens)
- Slim: https://www.anchorterminal.com/niches/code.understand-codebase/index.min.md (~280 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/code.understand-codebase/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 0 | 8 | 25 | 14 | 11 |

## What counts as done

Definitions, references or the relevant files are named with repository and path, and the answer is checkable against the current revision.

Measured, for our own tests: the answer comes in one or two calls, and the total tokens returned are lower than a grep of the same question would have cost.

How agents ask for it: where is this function used across our repos; give me an overview of this repository; find the code that handles X; what calls this before I change it

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Azure DevOps MCP Server (Microsoft Corporation)**: search_code is full-text search across the organisation and repo_file reads files; no definitions or references, and one account saw a project listing fill 160,000 tokens of context.
- **Entire (Entire Inc)**: The entire-graph plugin's graph query ranks source regions within a 24,576-byte budget and graph impact lists callers and callees with file and line; one repository per --repo, the plugin must be installed first, and the token saving is Entire's own figure.
- **GitHub MCP Server (GitHub, Inc.)**: search_code searches across every repository the token reaches and get_file_contents takes fields and ref; no definitions or references, and the token cost of the schema is 4 to 32 times a gh call.
- **GitLab MCP Server (GitLab Inc.)**: search and semantic_search run across projects and get_repository_file caps at 2,000 lines; one account relies on it for cross-repository reads, but there are no definitions or references.
- **GitMCP (Ido Salomon (GitMCP open-source project))**: search_<repo>_code uses GitHub code search within one public repository and one account saved hours browsing MLflow with it; one repository per URL, no private code, no definitions or references.
- **graft (Vishisht Choudhary (individual))**: /v1/grep, /v1/search and /v1/log run server-side without a clone and take limit and path; one repository per server, no definitions or references, and a project first committed on 13 October 2026.
- **Graphify (Graphify Labs)**: query_graph, get_neighbors, shortest_path and god_nodes answer within a 2,000-token default budget, with CLI and skill routes for 24 assistants; but one benchmark saw 5 per cent more tokens, large-repo queries were slow, and agents fall back to grep.
- **Sourcegraph MCP Server (Sourcegraph, Inc.)**: keyword_search and nls_search span the instance, go_to_definition and find_references give compiler-accurate results, read_file takes startLine and endLine; two accounts report better results and fewer tokens; but Enterprise only from $16,000 a year, and code_finder and deepsearch error when their quota is spent.

- not this job: amCharts 5 MCP Server (amCharts): Serves vendor documentation, not the user's repositories.
- not this job: Context7 (Upstash): Indexes documentation, including private sources on Pro, not code search over a repository.
- not this job: Git (MCP reference server) (Model Context Protocol, a Series of LF Projects, LLC): git_log, git_show and git_branch read one local repository; no search or navigation.
- not this job: Jam MCP (Jam): No code tools.
- not this job: Microsoft Learn MCP Server (Microsoft Corporation): Microsoft documentation only; no repository tools.
- not this job: ModusToolbox AI Assistant (Infineon Technologies AG): Analyses a ModusToolbox workspace for device context; not a code search tool.
- not this job: Momentic MCP server (Momentic): No code tools.
- not this job: REA (morluto): Analyses compiled binaries, Electron apps and websites without source; it does not search a source repository.
- not this job: Ref (Ref Software, Inc): Indexes documentation and PDFs; not a code search tool.
- not this job: Salesforce DX MCP Server (Salesforce, Inc.): No code search tools.
- not this job: Superpowers (Prime Radiant (no legal form or registration found on its pages); the MIT LICENSE names Jesse Vincent): No code search or graph.
- not this job: TokenEconomizer (Microsoft Corporation (repository in the microsoft GitHub organisation with Microsoft's SECURITY.md); co-authors at UC Berkeley): Injects learned facts about a repository; it does not search or map code.
- not this job: Void (Cloudflare, Inc. (VoidZero)): No code search.

## What stops agents finishing

- Enterprise plan from a $16,000 annual minimum (Sourcegraph)
- code_finder and Deep Search draw on a credit quota and error when it is spent
- graph build of 15 to 26 minutes on a large repo before the first query
- agents ignore the graph tool and fall back to grep unless the harness forces it

## Around it

Before: code.host-ops. After: code.ticket-to-pr, code.review-pr.

## Where this comes from

Weight 14 of 100 in Code & developer platforms: 11 of 65 accounts (Sourcegraph 2, GitLab cross-repo 1, Graphify 4, DeepWiki 3, GitMCP code search 1); several measure token use, which is why users adopt or drop these tools

As data: https://www.anchorterminal.com/api/v1/use-cases/code.understand-codebase.json, or the MCP tool get_use_case with id code.understand-codebase. Draft, read 2026-11-12 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
