Cody review

Cody sits between a company’s code repositories and the developers or agents that need to understand and change them. Sourcegraph presents it as a codebase context platform, not just a standalone chat interface: Code Search finds code across repositories, Deep Search answers natural-language questions with citations, and Diff Tour organizes related files for review. Batch Changes and Agentic Batch Changes address coordinated changes across large codebases.

In a stack, teams can use the Sourcegraph platform for code search and context, connect through its APIs, or use its CLI in a terminal. The homepage also lists MCP as code graph knowledge for agents. Enterprise is offered as single-tenant cloud. That is a cloud deployment option, not evidence of a self-hosted server. The platform is aimed at teams working across monorepos or multiple repositories, where engineers and coding agents need repository facts before deciding what to change. Its value depends on fitting those workflows and repository access patterns.

Sourcegraph lists APIs for integrating with its platform and a CLI for using Sourcegraph in a terminal. It also describes MCP as code graph knowledge for agents. The homepage does not name an API style, SDK, programming language or plugin; confirm SDK and language support in Sourcegraph’s documentation before planning an integration. Do not assume a REST API or a particular client library from the general APIs label.

The documentation is a listed product resource, alongside a changelog and security portal. The integration surface is therefore the APIs, CLI and MCP offering, with docs as the place to check supported operations, authentication and versioning. For an agent workflow, check how MCP exposes code graph context and what credentials and repository permissions it requires. For a production integration, ask Sourcegraph about API limits, compatibility guarantees and how API access relates to Enterprise seats and AI credits. Those details affect whether the CLI or API is the right fit for automation.

Sourcegraph lists one plan: Enterprise, priced per contract per year. The minimum annual contract is $16K; the vendor does not list a monthly price or a per-month annual equivalent. There is no free plan or free trial listed, so an individual developer cannot start on a published free tier and upgrade later.

Enterprise includes AI credits per user, with no monthly credit expiry, and single-tenant cloud. The price page does not publish the included credit quantity or a per-user price, so teams should ask how seat count and consumption affect the quote. Volume credit buckets are an add-on with unpublished pricing. Optional customer success manager and premium support also carry unpublished prices. The available pricing details do not establish an overage rate or what happens when included credits run out; get those terms in writing before forecasting a production bill. The $16K minimum makes Cody an enterprise procurement decision, not a low-cost API to add to a small project.

The vendor does not publish a licence or source-availability statement on the vendor's site available to buyers. We cannot treat Cody as an open-source project or assume that its server can be self-hosted. Sourcegraph lists single-tenant cloud for Enterprise, but that is a cloud deployment model rather than a self-hosting commitment. Ask the vendor which components, if any, can run in your environment and under what licence.

Data export and migration paths are also not published. Before connecting Cody to repositories or building agent workflows around its APIs, ask how to export configuration, search data and usage records, and whether code context can be reproduced with another provider. Code remains in the repositories you control, but integrations, permissions and agent behavior may still need work to replace. Confirm exit and retention terms in the contract.

Sourcegraph says its platform handles large monorepos and multi-repository architectures. Its homepage cites SOC 2 Type II and ISO 27001, zero data retention for LLM inference, and enterprise authentication with SSO (SAML, OpenID Connect and OAuth), SCIM and RBAC. Enterprise also includes dedicated support, described as account managers and support engineers, and single-tenant cloud. These are relevant controls, but teams should verify how each applies to their deployment and contract.

The public pricing details do not give API rate limits, credit quantities, overage rates or service-level commitments. Ask how API limits behave under agent workloads, what happens when credits are exhausted, and what support response terms apply. Confirm repository indexing and update behavior for your architecture, plus data export and retention settings. The vendor cites zero retention for LLM inference; clarify the scope of that statement and how it interacts with logs, search indexes and operational records before making Cody a production dependency.