Sourcegraph Cody

Read this part first

If you found this page looking for a free AI coding assistant, or a cheaper alternative to Copilot, or something to install this afternoon — Cody isn't it, and hasn't been since mid-2025.

Sourcegraph stopped accepting new Cody Free and Cody Pro signups on June 25, 2025. On July 23, 2025, it terminated every existing account on those plans. Not grandfathered, not migrated — ended. The Enterprise Starter plan also stopped including Cody.

Sourcegraph's own announcement was straightforward about why: simplifying the product line and concentrating effort where it saw developers succeeding. Pro subscribers weren't charged during the wind-down and were offered $40 in credits for Amp, the company's other product.

So: Cody in 2026 is an enterprise product with no individual tier at any price. If you're a solo developer, skip to the alternatives section. If you're a tech lead sizing up tooling for fifty engineers across forty repos, keep reading, because Cody does something the popular tools genuinely don't.

What Cody is good at

One thing, done better than anyone: cross-repository context.

Cody isn't an AI assistant with a codebase index bolted on. It's a layer over Sourcegraph's code intelligence backend — the same infrastructure that indexes symbol definitions and usages across hundreds of repositories for code search. That means when Cody answers a question, it can reason about architectural dependencies spanning your entire organization, not just the files in your editor.

For a company with a sprawling multi-repo estate, that's a different category of understanding. Copilot sees your open files. Cursor indexes your project. Cody sees the whole estate, including the service three teams over whose interface you're about to break.

Sourcegraph has been doing this since 2013 and serves over 250,000 repositories, with customers including Dropbox, Stripe, and Canva. The code graph is mature technology, not a recent feature.

Supporting capabilities: a large context window, Context Filters for controlling what the AI can see, admin-configurable model selection across Anthropic, OpenAI, and others, and deployment options spanning cloud, self-hosted, and air-gapped.

The technical caveat worth knowing

Cross-repo navigation is the selling point, so it's worth understanding how it actually resolves — and where it doesn't.

Sourcegraph indexes each repository independently. Cross-repo references are joined at query time on package identity, matching a reference in one repo to a definition in another. That join succeeds when both sides are indexed at matching package versions.

When versions differ, or one side isn't indexed, precise navigation isn't available. Sourcegraph falls back to search-based fuzzy results.

The important part: the fallback is quiet. You still get an answer. It just comes from text search rather than a resolved symbol, and nothing tells you the difference. In a well-maintained monorepo this rarely bites. In a heterogeneous estate with drifting dependency versions — which is most large organizations — it happens more than you'd expect.

Worth asking Sourcegraph about index coverage against your actual repository set during evaluation, rather than assuming precision by default. (This detail comes from a competing vendor's analysis, so verify it against Sourcegraph's own precise code navigation docs — but the mechanism it describes is documented behavior, not marketing.)

What it costs

No self-serve tier, no published per-seat rate, sales-led contracts. Two figures circulate:

• Cody Enterprise — around $59 per user per month on annual contract, with some sources reporting closer to $49 depending on package
• Sourcegraph Enterprise platform — starting near $16,000, scaling with team size and AI-feature credits, bundling Cody's AI with code search, Deep Search, batch changes, MCP/API/CLI access, and admin controls

The $16,000 framing is how Sourcegraph presents it publicly. The per-seat figure is what buyers report.

For comparison, $59 per developer per month is roughly six times GitHub Copilot Pro. That's the number to justify, and the justification has to be something Copilot can't do at all — not "somewhat better completions." Cross-repo architectural context is that thing, but only if your codebase is genuinely sprawling enough to need it.

There's also a procurement argument that matters more than it sounds: one contract covers code search and AI together, rather than separate subscriptions for each. For organizations already paying for Sourcegraph, the marginal AI cost looks very different from the standalone figure.

The Amp question

Sourcegraph now runs two coding products, and the relationship creates real uncertainty.

Cody is the enterprise assistant on the platform. Amp is the newer agentic tool — multi-step task automation, Smart Mode on current frontier models, code graph context, agentic code review — and it's where Sourcegraph pointed former Cody Free and Pro users in 2025.

Multiple 2026 analyses observe that investment is shifting toward Amp, leaving Cody's long-term portfolio position unclear. Nobody's announced anything, and Sourcegraph explicitly reaffirmed its commitment to Cody Enterprise during the 2025 changes. But if you're signing a multi-year enterprise contract, the split is worth raising directly with sales.

And if what you actually want is agentic behavior rather than context-aware assistance, evaluate Amp on its own terms rather than assuming Cody is the flagship.

What individual developers should use instead

Since Cody closed that door, briefly:

• GitHub Copilot at $10/month — the sensible default, predictable, works everywhere
• Cursor at $20 — the deepest AI-native IDE experience with mature agents
• Cline — free and open source, bring your own API key, pay only for inference
• Augment Code — if cross-repo context specifically is your problem and you want it without an enterprise contract
• Tabnine — local, privacy-first completion running on-device
• Amp — Sourcegraph's own answer, if you want the code graph without the enterprise commitment

Cody vs the enterprise alternatives

Against Augment Code: Augment indexes 400,000-plus files and now offers its Context Engine as an MCP server, so other agents can use its indexing. It's credit-priced from $20 and available without a sales cycle. Cody has deeper code-search heritage and stronger enterprise deployment options including air-gapped. For context at scale without enterprise procurement, Augment; for a single vendor covering search and AI with full governance, Cody.

Against Cursor: Cursor is a better daily editing experience with more mature agents at $20 per seat. It indexes your project well but doesn't reach across an organization's full repository estate. For developer experience, Cursor; for cross-repo architecture, Cody.

Against Copilot Enterprise: GitHub Copilot has the ecosystem, the price, and native GitHub integration. Cody has meaningfully better multi-repo context. If your code lives in GitHub and your repos are manageable, Copilot; if repository sprawl is the actual problem you're solving, Cody.

Our Verdict

Cody does one thing better than any competitor, and it's a real thing: understanding a large multi-repository codebase as a whole rather than as a collection of files. Built on thirteen years of Sourcegraph code intelligence, with air-gapped deployment, admin-configurable models, and Context Filters, it's a genuine enterprise product serving genuine enterprise problems. For an organization where onboarding takes weeks because nobody can trace a dependency across forty repos, that's worth real money.

The reservations are about access and direction. Killing Free and Pro in July 2025 didn't just remove a cheap tier — it removed the evaluation path, so teams now commit to enterprise conversations before testing at meaningful scale. At roughly $59 per developer it's six times Copilot, which demands a specific justification most codebases can't provide. Cross-repo precision depends on index coverage and version matching, with a quiet fuzzy fallback when those conditions aren't met. And with Sourcegraph's investment visibly moving toward Amp, Cody's roadmap position deserves a direct question before a multi-year signature.

For large engineering organizations where repository sprawl is the actual bottleneck, worth evaluating — and ask about index coverage and the Amp roadmap. For everyone else, Cody stopped being an option twelve months ago.

Note: Sourcegraph does not currently have an active affiliate program with AIVario. We earn no commission, and this rating carries no commercial incentive.

Best for: Large engineering organizations with hundreds of repositories, teams where cross-repo dependency tracing is a real cost, companies wanting one vendor for code search and AI, orgs needing air-gapped or self-hosted deployment Not ideal for: Individual developers (no tier exists), small teams on a handful of repos, anyone wanting to evaluate before a sales conversation, teams primarily wanting agentic coding (look at Amp) Bottom line: The best cross-repo context in AI coding, sold only to enterprises since it deleted its free tier in July 2025. If your repos sprawl, evaluate it; if they don't, Copilot at a sixth the price does the job.

Related Tools

• GitHub Copilot — $10/month default for individuals and most teams
• Cursor — deepest AI-native IDE with mature agent workflows
• Augment Code — cross-repo context without an enterprise contract
• Cline — free open-source agent, bring your own model
• Tabnine — local, privacy-first completion

Frequently Asked Questions about Sourcegraph Cody

Is Sourcegraph Cody still free?

No. Sourcegraph stopped accepting new Cody Free and Cody Pro signups on June 25, 2025, and terminated all existing accounts on those plans on July 23, 2025. The Enterprise Starter plan also stopped including Cody. Existing Cody Enterprise customers were unaffected. Pro subscribers weren't charged during the transition and were offered $40 in credits for Amp, Sourcegraph's other product, as a thank-you. If you're reading an older guide that mentions a free Cody tier, it's describing something that no longer exists.

How much does Cody cost now?

Two figures circulate and both appear valid depending on packaging. Cody Enterprise is reported around $59 per user per month on an annual contract, with some sources citing closer to $49. Separately, Sourcegraph's Enterprise platform — which bundles Cody's AI with code search, Deep Search, batch changes, and admin controls — starts near $16,000 and scales with team size and AI-feature credits. There's no self-serve tier and no published per-seat rate; expect a sales-led contract.

What should individual developers use instead?

Sourcegraph's own answer is Amp, its agentic coding product, which is where it pointed former Cody Free and Pro users during the 2025 transition. Amp brings the code graph context, multi-step agent automation, and a Smart Mode running current frontier models. Beyond Sourcegraph, GitHub Copilot at $10/month is the sensible default, Cursor is the deepest AI-native IDE, Claude Code handles terminal agentic work, and Cline gives you a free open-source agent with your own API key.

What is Cody actually best at?

Cross-repository context at scale, and it's genuinely unmatched there. Cody sits on Sourcegraph's code intelligence backend, which indexes symbol definitions and usages across hundreds of repositories — so the AI understands architectural dependencies spanning your whole organization rather than the files you have open. For a company with a sprawling multi-repo estate, that's a category of understanding tools like Copilot and Cursor don't reach.

Does cross-repo navigation always work?

Not always, and the failure mode is quiet. Sourcegraph indexes each repository independently, then joins a reference to a definition on package identity at query time. That join lands only when both sides are indexed at matching package versions. When versions differ or an index is missing, precise navigation isn't available and Sourcegraph falls back to search-based fuzzy results. You still get an answer — it just comes from text search rather than a resolved symbol, without announcing the downgrade. Worth knowing before you assume precision.

Which models does Cody use?

Admin-configurable rather than fixed. Sourcegraph's model configuration supports provider overrides, model overrides, self-hosted models, defaults, and filters. Cody Gateway gives Enterprise customers zero-configuration access to models from providers including Anthropic and OpenAI, while self-hosted deployments can point at third-party LLM providers directly. That flexibility matters for organizations with existing model commitments or data-residency requirements.

How does Cody relate to Amp?

They're sibling products with a split in buyer intent, and the relationship creates genuine roadmap uncertainty. Cody is the enterprise assistant layered on Sourcegraph's platform; Amp is the newer agentic tool aimed at individual developers, with multi-step task automation and agentic code review. Multiple 2026 analyses note Sourcegraph's investment shifting toward Amp, leaving Cody's long-term position within the portfolio unclear. If agent behavior is what you want, evaluate Amp directly rather than assuming Cody is the flagship.

Is Cody worth $59 per developer?

It depends entirely on repository sprawl. At roughly six times GitHub Copilot Pro's price, the justification has to come from something Copilot can't do — and cross-repo architectural context genuinely is that thing at real scale. For a 50-developer organization spanning dozens of repos where onboarding and dependency tracing eat real time, the math can work. For a team on a handful of repos, you're paying enterprise rates for capability your codebase doesn't require.