Quick Verdict
Sourcegraph Cody is the AI coding assistant within the Sourcegraph Enterprise platform. Its differentiator is not a low-cost individual autocomplete subscription but Sourcegraph Search context across large, multi-repository codebases. Developers can use supported IDE integrations, the web interface, and CLI workflows for code questions, explanations, generation, and changes. Legacy individual Cody plans should not drive current procurement. The Sourcegraph platform starts at approximately $16,000, with final pricing depending on deployment, users, scale, and contract.
Cody fits organizations that already need enterprise code search, permission synchronization, and deployment controls. Sourcegraph offers cloud and self-hosted routes and states that customer code and prompts are not used to train generative models. That no-training commitment does not remove model-provider processing, logs, retention, or configuration-specific data flows. The product deserves an enterprise proof of concept rather than a casual individual-tool recommendation.
Best For
- Medium and large engineering organizations with many private repositories and services.
- Teams already purchasing or evaluating Sourcegraph Enterprise Search.
- Developers who need permission-aware code context in IDE, web, and CLI workflows.
- Architecture, onboarding, incident-response, and legacy-code investigation teams.
- Not suitable for individual developers, small repositories, or budgets far below the enterprise platform entry price.
Key Features
- Search context: finds symbols, references, files, and paths across repositories before constructing an answer.
- IDE use: supports code questions, explanations, generation, and changes in supported editors.
- Web interface: investigates indexed code without opening every repository locally.
- CLI workflow: brings Cody capabilities to terminal and automation scenarios under current enterprise controls.
- Enterprise permissions: combines code-host connections and permissions so retrieved context should follow user access.
- Cloud or self-hosted: offers deployment choices for compliance, operations, and network requirements.
- No-training commitment: Sourcegraph states that customer code and prompts are not used to train generative models.
Use Cases
In a microservice estate, Cody can investigate how an identity field travels from APIs to storage by retrieving context from several repositories. New hires can ask about module responsibilities in an IDE or on the web and then verify cited source files. During an incident, it can help find call sites and configurations but cannot replace runtime evidence, logs, or accountable engineering judgment.
A serious pilot should select a cross-repository question with a known human answer and measure coverage, citation traceability, latency, and errors. Use two test accounts with different permissions to verify isolation. Compare Sourcegraph Cloud with self-hosting for operations, model routing, and data residency.
Pricing
Current Cody should be treated as a Sourcegraph Enterprise platform capability, not a standalone freemium personal product. Public platform pricing starts around $16,000, while the final quote varies with users, code scale, deployment, support, and contract term. There is no stable standalone Cody price suitable for direct comparison with a $20 individual coding subscription.
Total cost also includes code-host connections, indexing resources, permission synchronization, model usage, upgrades, monitoring, and self-hosting operations. Cloud reduces infrastructure work but places service metadata and relevant code context in a supplier environment. Self-hosting increases control and capacity responsibilities. Model-provider costs and terms belong in the same quote review.
Pros
- Search context is designed for large, multi-repository codebases.
- IDE, web, and CLI cover different engineering workflows.
- Integrates with enterprise code hosts, indexing, and permissions.
- Supports Sourcegraph Cloud and self-hosted deployment choices.
- Explicitly states that customer code and prompts are not used for generative-model training.
Cons
- Platform entry price is around $16,000, excluding most individuals and small teams.
- Cody is not currently a low-cost standalone subscription.
- Deployment, indexing, permissions, and model routing require administration.
- No-training does not mean zero retention or no external model processing.
- A self-hosted platform can still send context to an external model endpoint.
- Generated answers and changes still require source, test, and review verification.
Alternatives
| Tool | Best for | Advantage | Limitation |
|---|---|---|---|
| Amp | Individuals and teams wanting a cloud coding agent | Lower subscription and usage entry points, direct task execution | No self-hosting and not an enterprise Search platform |
| GitHub Copilot | GitHub and mainstream IDE users | Mature procurement and broad editor integration | Different cross-code-host Search focus |
| Cursor | Developers preferring an AI-native IDE | Intuitive multi-file and agent workflow | Different enterprise search-platform depth |
| Claude Code | Terminal-heavy developers | Direct investigation, editing, and commands | Enterprise-scale indexing and permissions need separate systems |
FAQ
Is Cody still a standalone personal AI coding product?
Current Cody should be understood as part of Sourcegraph Enterprise. Legacy individual plans and prices should not be used for current purchasing decisions.
How much does Sourcegraph Cody cost?
The public platform entry point is approximately $16,000, but Cody has no stable standalone price. Obtain a written quote for the actual scale and deployment.
What is the difference between Cody and Amp?
Cody uses Sourcegraph Search and enterprise platform context. Amp is a separate proprietary cloud coding agent with $20/$200 subscriptions and pay-as-you-go billing.
Does Sourcegraph train models on private code?
Sourcegraph states that customer code and prompts are not used to train generative models. Model-provider processing, retention, logs, and contracts still require review.
Does self-hosting guarantee code stays inside the network?
No. Indexing and the application can be self-hosted, but context can still leave the network if Cody uses an external model endpoint.
Where can developers use Cody?
The current product covers supported IDE integrations, Sourcegraph web, and CLI scenarios. Exact clients and features depend on the enterprise version.
Bottom Line
Sourcegraph Cody is best understood as Search-context AI for large enterprise code estates, not as a low-cost personal coding-agent subscription. IDE, web, CLI, and cloud/self-hosted choices fit complex organizations, while the roughly $16,000 platform starting point requires measurable proof of value. Validate permission isolation, search citations, model data flow, and no-training terms, then compare the enterprise outcome with Amp and GitHub Copilot.