Cody Review (2026): Sourcegraph's Enterprise Codebase-Aware Coding Assistant

Cody Review (2026): Sourcegraph's Enterprise Codebase-Aware Coding Assistant

Cody is Sourcegraph's AI coding assistant, built on top of the company's long-standing code search platform. Its defining idea has always been context: rather than reasoning about a single open file, Cody retrieves relevant code from across an entire codebase, and in enterprise settings across many repositories at once, before it answers a question or writes code. That codebase-awareness is what set Cody apart from editor-native autocomplete tools, and it remains the reason large engineering organizations still reach for it.

One important caveat frames this whole review: Cody's availability changed significantly in 2025. This hand-reviewed update covers what Cody is in August 2026, who can still get it, how pricing works, its codebase-context strengths, and how it stacks up against GitHub Copilot, Cursor, and Claude Code.

Is Cody still available, and for whom?

This is the first thing to settle, because the answer is not what it was a couple of years ago. On June 25, 2025, Sourcegraph announced it would discontinue the Cody Free and Cody Pro plans as of July 23, 2025, and that the Enterprise Starter plan would no longer include Cody. Individual developers and small teams who relied on the free or low-cost tiers lost access on that date.

What remains is Cody Enterprise, which Sourcegraph has said is unaffected by those changes and continues to be fully supported and actively developed. In practical terms, as of 2026 Cody is best understood as an enterprise-only product aimed at larger, security-conscious organizations rather than something an individual can sign up for on a credit card.

Sourcegraph pointed former Free and Pro users toward Amp, a separate agentic coding tool it introduced for individuals and teams. Reporting through late 2025 indicates Amp was subsequently spun out as an independent company, distinct from Sourcegraph and Cody. That corporate detail is worth treating as approximate rather than settled, but the practical takeaway is clear: if you are an indie developer looking for a personal Cody subscription, that path effectively closed in mid-2025, and Amp (or a rival tool) is the intended alternative.

At a glance

Pricing

Because Cody is now sold as part of Sourcegraph's enterprise platform, there is no published per-seat rate card the way there is for consumer tools. Sourcegraph's own materials have indicated enterprise pricing that starts in the region of $16,000 for a contract, bundling AI feature credits that scale with team size, and final pricing is negotiated per customer.

Third-party review sites through 2026 have circulated per-seat figures for Cody Enterprise in the range of roughly $49 to $59 per user per month on an annual commitment, sometimes higher. These numbers should be treated cautiously: they come from secondary sources rather than a public Sourcegraph price list, and actual cost depends on seat count, deployment model, and negotiated terms. The safe summary is that Cody is priced as enterprise software, typically involving an annual contract and a conversation with sales, not a self-serve checkout.

One thing that has genuinely gone away is the cheap on-ramp. The former Cody Free (basic completions and chat at no cost) and Cody Pro (around $9 per month for higher limits) plans no longer exist, so there is no low-cost individual tier to evaluate the tool on your own.

Key features

Cody's feature set is built around three surfaces that most AI coding assistants share, differentiated by the depth of context feeding each one:

• Codebase-aware chat. You can ask natural-language questions about how a system works, where a function is used, or why a pattern exists, and Cody answers using code retrieved from your actual repositories rather than generic training data. This is the feature most tied to Sourcegraph's search heritage.
• Inline autocomplete. Cody offers single- and multi-line completions in the editor as you type, drawing on surrounding code and, in enterprise setups, wider repository context.
• Commands and prompts. Prebuilt and custom commands handle recurring tasks such as explaining a selection, generating unit tests, writing documentation, or drafting a fix, which standardizes common workflows across a team.
• Model choice. Cody Enterprise lets administrators select which frontier model backs the assistant, with options spanning Anthropic's Claude, OpenAI's GPT, and Google's Gemini families. This avoids locking an organization to a single vendor and lets teams tune for cost or capability. The exact model versions available shift over time, so treat any specific version name as a moving target.
• Enterprise controls and deployment. Cody supports cloud, single-tenant, and self-hosted or air-gapped deployments, with SOC 2 compliance and Sourcegraph's stated guarantees that customer code is not used to train models. These are the features that justify the product's positioning for regulated and security-sensitive organizations.

Sourcegraph has also continued to invest in more autonomous, agentic behavior within its ecosystem, though the most explicitly agent-first work has centered on Amp. Cody's own emphasis remains context-rich assistance inside the developer's editor.

Codebase context: the core strength

If there is a single reason to choose Cody, it is context on large codebases. Sourcegraph indexes your repositories and uses that index, alongside embeddings, to fetch the most relevant snippets from across a codebase, and in enterprise deployments across many repositories at once, before sending a prompt to the model. For a monorepo or a sprawling microservice estate, this cross-repository retrieval is meaningfully different from tools that primarily see the open file and a few neighbors.

In practice this helps most with questions that span services: tracing how a request flows through several repos, finding every caller of a shared library, or understanding an unfamiliar subsystem before changing it. For organizations whose pain is code comprehension at scale rather than greenfield generation, this is Cody's differentiator and the capability its enterprise pricing is built around.

Strengths and limits

The clearest strength is that codebase-scale context, backed by a mature search engine rather than bolted on as an afterthought. Combined with model choice, flexible deployment including air-gapped options, and enterprise compliance posture, Cody fits organizations that cannot send code to a black-box consumer service and that need answers grounded in their real, large repositories.

The limits are mostly about access and momentum. The removal of Free and Pro plans means there is no easy way for an individual to try Cody or advocate for it bottom-up, which historically is how developer tools spread inside companies. Adoption now depends on an enterprise decision. There is also a competitive question: much of the industry's energy in 2025 and 2026 shifted toward autonomous, agentic coding, and Sourcegraph itself channeled that ambition into the separate Amp product. That leaves Cody positioned more as a context-rich in-editor assistant than a cutting-edge autonomous agent, which may or may not match where a given team wants to invest. As with any LLM tool, generated code and answers still require human review.

Pros and cons

Pros

• Best-in-class codebase context, including cross-repository retrieval on large monorepos and microservice estates
• Built on Sourcegraph's mature code search index rather than a lightweight add-on
• Admin-selectable models across Anthropic, OpenAI, and Google, avoiding single-vendor lock-in
• Flexible deployment: cloud, single-tenant, and self-hosted / air-gapped
• Strong enterprise posture: SOC 2 and stated no-training guarantees on customer code

Cons

• Enterprise-only since mid-2025; no free or low-cost individual tier remains
• Pricing is contract-based and not transparently published, requiring a sales conversation
• Individuals cannot easily trial it or drive bottom-up adoption
• Positioned around in-editor assistance rather than the autonomous, agentic workflows Sourcegraph moved into Amp
• Output still needs human verification

How it compares

GitHub Copilot is the most direct point of comparison for teams and the lowest-friction option, with per-seat plans, deep VS Code and JetBrains integration, and broad organizational availability. Copilot is far easier to adopt individually or team-wide, but its context has historically centered on the working file and repository rather than the cross-repository, index-backed retrieval that is Cody's specialty. For a large multi-repo estate where code comprehension is the bottleneck, Cody's context can be the deciding factor; for straightforward per-seat rollout and inline autocomplete, Copilot is simpler. See our Cody vs GitHub Copilot comparison for a closer look.

Cursor is an AI-first editor with a polished graphical experience, strong agentic features, and a large community. It is aimed squarely at individual developers and teams who want an integrated IDE, and it is trivially easy to try, which is exactly where Cody no longer competes. Cursor emphasizes an in-editor agentic workflow; Cody emphasizes retrieval-grounded answers across big codebases and enterprise deployment control.

Claude Code, from Anthropic, sits at the terminal-first, agentic end of the spectrum, taking a goal and editing across files, running commands, and self-correcting. It is a different shape of tool from Cody: an autonomous collaborator you supervise, versus a context engine embedded in your editor. Notably, this is closer to the territory Sourcegraph steered toward with Amp than with Cody itself.

Who it's for

Cody in 2026 is for large, security-conscious engineering organizations with big or multi-repository codebases, especially those that need self-hosted or air-gapped deployment and strong compliance guarantees. If your primary pain is understanding and safely changing a sprawling codebase, and you can commit to an enterprise contract, Cody's context capabilities are a genuine fit.

It is not a fit for individual developers, hobbyists, or small teams looking to sign up and pay monthly on their own; that door closed in mid-2025. Those users are better served by Copilot, Cursor, Claude Code, or Sourcegraph's own Amp.

Verdict

Cody remains one of the strongest tools available for understanding large codebases, thanks to Sourcegraph's search-backed, cross-repository context and its flexible, compliance-friendly deployment options. But the product has narrowed sharply. With Free and Pro discontinued in July 2025, Cody is now an enterprise-only offering, sold through contracts rather than self-serve plans, while Sourcegraph directed its agentic ambitions into the separate Amp product. For big organizations that need grounded answers across many repositories and control over where their code runs, Cody is well worth evaluating. Everyone else should look at Copilot, Cursor, or Claude Code, since the individual on-ramp that once made Cody easy to try no longer exists.

FAQ

Is Cody still available in 2026? Yes, but as an enterprise-only product. Sourcegraph discontinued the Cody Free and Cody Pro plans on July 23, 2025, and Enterprise Starter no longer includes Cody. Cody Enterprise remains supported and actively developed.

Can individuals still use Cody for free? No. The free and low-cost individual tiers were removed in mid-2025. Sourcegraph pointed former users toward its separate Amp product; other alternatives include GitHub Copilot, Cursor, and Claude Code.

How much does Cody Enterprise cost? Pricing is contract-based and negotiated per organization. Sourcegraph has indicated enterprise pricing starting in the region of $16,000, and third-party sources cite roughly $49 to $59 per user per month on annual terms, figures that should be treated as approximate rather than official.

What makes Cody different from GitHub Copilot? Cody's cross-repository context, powered by Sourcegraph's code search index, is built for reasoning across large multi-repo codebases, whereas Copilot centers more on the working file and repository and is far easier to adopt per seat. See our Cody vs GitHub Copilot comparison.

Which AI models does Cody use? Cody Enterprise lets administrators choose among frontier models from Anthropic, OpenAI, and Google. The specific versions available change over time, so treat any single model name as a moving target.