ToolWorthy
Sourcegraph icon

Sourcegraph

Sourcegraph is enterprise code intelligence and context infrastructure for cross-repository search, navigation, investigation, monitoring, large-scale changes, and permission-aware AI agent access.

Content updated 1 month ago

Pricing:Free + Premium
Get Started
Sourcegraph enterprise code intelligence platform

Verdict

Sourcegraph is enterprise code intelligence and context infrastructure, not simply an AI code generator. It indexes code across repositories and hosts so developers and agents can search, navigate, investigate, monitor, and change large codebases with organization-wide context and repository permissions.

It is most defensible for a large or fragmented code estate where cross-repository understanding, governed agent access, deployment control, and repeatable migrations matter. A small team working in a few local repositories may get enough direct coding help from Cursor, Claude Code, or Codex without buying an additional enterprise context layer.

Search Jobs

  • What Sourcegraph is now and how it differs from a coding assistant
  • Sourcegraph Free, Enterprise Starter, Dedicated Cloud, and Self Hosted pricing
  • Sourcegraph MCP for Claude Code, Cursor, Codex, and other agents
  • Sourcegraph cloud versus self-hosted deployment
  • Code Search, Deep Search, Batch Changes, Insights, and Monitoring
  • Current boundaries between Sourcegraph, Cody, and Amp

What Sourcegraph Actually Is

Sourcegraph connects to repositories on major code hosts and builds an indexed, permission-aware view of the code estate. Humans use that layer through web and IDE experiences; agents can reach it through MCP, APIs, or other integrations. The product's value is not primarily generating a function from a prompt. It is making the wider codebase discoverable and governable before a human or agent proposes a change.

The main capability groups solve different jobs:

  • Code Search finds symbols, strings, patterns, commits, and code across repositories and hosts.
  • Code Navigation follows definitions and references to understand unfamiliar code paths.
  • Deep Search investigates a question across indexed code and returns an evidence trail rather than a single lexical match.
  • Batch Changes plans and manages repeatable changes across many repositories.
  • Insights tracks code patterns or migration progress over time.
  • Monitoring watches for code changes that match important queries.

How Agents Use Sourcegraph Context

Sourcegraph's MCP server lets supported coding clients ask search and context questions against the enterprise instance. Current official materials list Claude Code, Cursor, Codex, Amp, and other MCP-capable clients. The MCP route is an integration surface: the coding agent remains separate, while Sourcegraph supplies indexed code context under Sourcegraph's repository permissions.

This is the distinctive complement for enterprise agents. An IDE assistant can reason well about files already open or locally available; Sourcegraph is designed to help it locate relevant code across a much larger estate and to preserve organization-level access controls.

MCP is currently an Enterprise capability. Administrators control access, and clients authenticate through the documented OAuth or token paths. GraphQL and REST APIs and the Sourcegraph CLI provide additional automation routes for search and platform workflows.

Code Hosts and Deployment

Sourcegraph supports major code hosts and direct Git repository connections, including common GitHub, GitLab, Bitbucket, and Perforce environments. The current Enterprise plan supports single-tenant cloud and self-hosted deployment.

Single-tenant cloud reduces platform operations while preserving a dedicated service boundary. Self-hosting offers more infrastructure control but transfers upgrades, capacity, availability, backups, and incident response to the customer. The right choice depends on code-residency policy, network topology, procurement, and the internal team available to operate the system.

Security and Privacy Boundary

Repository permissions are central to Sourcegraph's design and are enforced when code context is retrieved through MCP. Enterprise buyers should still validate identity-provider integration, permission synchronization, audit requirements, network access, encryption, retention, and the chosen model-provider path.

Sourcegraph's Cody FAQ says enterprise customer data is not used to train shared models and that third-party model providers do not train on that data. Self-hosting Sourcegraph does not automatically mean every AI request stays on-premises: depending on configuration, code snippets can still be sent to a configured external model service. The model gateway and data-flow design therefore need explicit review.

Sourcegraph, Cody, and Amp

Sourcegraph is the enterprise code intelligence platform and context layer. Cody Enterprise remains the Sourcegraph-supported AI assistant experience across supported IDEs and the web, using Sourcegraph context. Amp is a separate coding-agent company after the 2025 separation announcement and should not be treated as the current name for Sourcegraph or as a bundled Sourcegraph plan.

Older Cody Free or Pro pages can describe a historical individual-assistant product shape. Current Sourcegraph evaluation should start from the Enterprise platform and current Cody Enterprise documentation, not stale per-seat assumptions.

Pricing Mechanics

Sourcegraph has a Free plan at no charge. It is the correct product entry point and is why this page is classified as freemium rather than paid-only.

Enterprise Starter is USD 19 per user per month and is positioned for teams of up to 50 developers.

Enterprise Dedicated Cloud is USD 59 per user per month and starts at 25 developers, so the published seat minimum implies a minimum of 25 paid seats for that deployment.

Enterprise Self Hosted is sold through a contract rather than a universal public self-serve checkout price. Buyers should confirm the current seat minimum, deployment terms, support, and AI-credit allowance in the quote; this page does not convert that contract boundary into a fabricated global entry price.

Paid enterprise economics therefore depend on per-seat price, the plan's minimum or maximum team-size rule, deployment, expected AI usage, support, and negotiated contract terms. A team should have an organization-level code-intelligence problem large enough to justify that layer, not only a desire for another coding chat window.

Best For

  • Enterprises with hundreds or thousands of developers and code spread across many repositories or hosts.
  • Platform teams governing large migrations, framework adoption, deprecations, or security changes.
  • Organizations that want permission-aware cross-repository context for multiple coding agents.
  • Teams needing self-hosted or single-tenant deployment and enterprise administration around code intelligence.

Avoid If

  • The team is small, its repositories fit comfortably in local tooling, and an IDE-first agent already finds enough context.
  • The only job is generating or editing code in the current repository; Sourcegraph's indexing and governance layer may be unnecessary.
  • The applicable seat minimum or enterprise contract is disproportionate to the codebase or expected adoption.
  • The organization cannot operate self-hosting and cannot accept the available cloud or external-model data flows.
  • No team owns repository connections, permissions, indexing health, search quality, and change-management workflows.

Sources & Verification

Track Sourcegraph in ToolWorthy Weekly

Important tool updates, better alternatives, and selected AI signals in one weekly brief.

Weekly only. Unsubscribe anytime.