Cody Is No Longer a Consumer Coding Assistant
Sourcegraph moved Cody firmly into enterprise
By WhatAI Editorial ·
Cody Is No Longer a Consumer Coding Assistant
For several years, Sourcegraph Cody was easy to describe. It was an AI coding assistant with an unusually strong understanding of large repositories. Developers could install an extension, chat with their codebase, receive completions, ask for edits, and draw context from Sourcegraph's code search. There was a free plan, a paid individual plan, and an enterprise version. That description is now historically accurate and commercially misleading.
Cody Free and Cody Pro were discontinued on July 23, 2025. Cody was also removed from Enterprise Starter. Sourcegraph directed individual users toward Amp, the agentic coding product created by Sourcegraph's founders. Cody Enterprise was not discontinued. It remains supported, appears in current documentation, and continued to receive model updates in August 2026. The difference is that Cody is now firmly an enterprise capability inside a much broader Sourcegraph platform, not a standalone freemium assistant competing for every individual developer.
This distinction matters for anyone landing on an old review or pricing table. A $9 or $19 Cody plan is no longer available. The current Sourcegraph pricing page starts at $16,000 for an enterprise platform that combines code search, Deep Search, Batch Changes, Insights, Monitoring, MCP access, APIs, CLI access, private deployment choices, enterprise controls, and credits for AI features. Cody still matters within that environment, especially in supported IDEs, but it should be evaluated as part of an enterprise code-understanding strategy.
What Still Exists
Sourcegraph's current Cody documentation describes an AI coding assistant available on Sourcegraph Enterprise through Visual Studio Code, JetBrains IDEs, Visual Studio, and the web application. Its core capabilities remain recognizable: developer chat, code completions, code edits, customizable prompts, model choice, and codebase context. Cody uses Sourcegraph's Search API to retrieve relevant information from local and remote repositories, including symbols, APIs, definitions, usage patterns, and files beyond the one currently open.
That context layer is still Cody's strongest idea. A generic assistant may perform well when the answer fits inside an open file or a small repository. Enterprise work often spans hundreds of repositories, shared libraries, duplicated services, generated code, ownership boundaries, and years of history. A developer asking where an interface is implemented may need an answer across several code hosts. A refactor may depend on usages in projects the developer did not know existed. Sourcegraph was built for that scale before coding assistants became popular, and Cody can use that existing search and indexing foundation.
Current product updates confirm that Cody has not been abandoned. Sourcegraph added support for the GPT-5.6 model family in its August 10, 2026 update. Recent releases also contain Cody-specific administration and compatibility work. At the same time, Sourcegraph 7.0 makes the hierarchy clear: when Deep Search is enabled, it replaces Cody Web in the Sourcegraph interface. The Cody web route and sidebar are hidden, while Cody's IDE experience remains a supported enterprise feature.
The Product Around Cody Has Become More Important Than Cody
Sourcegraph now describes itself as an intelligence layer for AI coding agents and developers. The current platform is organized around understanding, oversight, and change at codebase scale. Code Search finds exact patterns and symbols. Deep Search investigates open-ended questions with AI. The MCP server supplies Sourcegraph context to external agents such as Claude Code, Cursor, Codex, and Amp. Agentic Batch Changes applies migrations across many repositories. Insights and Monitoring help teams track patterns and risks over time.
In that portfolio, Cody is the in-IDE assistant. It gives developers a familiar chat, completion, and edit experience backed by Sourcegraph context. Deep Search is the research surface for questions that require broader investigation. MCP lets teams carry the same code intelligence into whichever coding agent they already prefer. Batch Changes handles coordinated work at a scale no single IDE session should own. This is a more coherent enterprise story than asking Cody to be every type of AI development tool at once.
It also changes the purchasing question. An engineering leader should not ask whether Cody produces better autocomplete than a cheaper assistant. The better question is whether Sourcegraph can become a reliable context and change layer across the organization's code estate. If the team buys Sourcegraph for search, navigation, security investigation, migrations, ownership, and agent context, Cody can be a useful interface on top. If the team wants only personal completions and chat, the current enterprise platform is probably excessive.
Cody and Amp Are Not the Same Product
The transition became more confusing when Sourcegraph pointed former Cody Free and Pro users toward Amp. Amp was initially developed alongside Sourcegraph, but Sourcegraph and Amp announced in December 2025 that they were becoming independent companies. Sourcegraph is now led as the code search and code understanding business, while Amp Inc. focuses on frontier coding agents. The shared history remains, but a buyer should not treat Amp as a new Cody plan or assume an Amp subscription grants Sourcegraph Enterprise capabilities.
The products now solve different problems. Amp is an agentic coding environment for developers and teams that want an AI to perform substantial implementation work. Cody Enterprise is an assistant integrated with Sourcegraph context. The broader Sourcegraph platform is designed to help humans and many different agents understand and evolve large codebases. Sourcegraph's MCP strategy is deliberately agent-agnostic: the company wants Claude Code, Cursor, Codex, Amp, and other tools to retrieve better context from Sourcegraph.
That separation can benefit customers. An organization may standardize on an external coding agent while using Sourcegraph for cross-repository retrieval, permissions, auditability, search, and large-scale change. Cody can remain available for developers who prefer an in-IDE assistant without forcing the whole organization into one agent. The downside is a more complex product map. Old Cody comparisons often combine capabilities that now belong to Cody, Deep Search, MCP, Batch Changes, or Amp. Procurement should name the exact workflow and SKU behind every requirement.
Privacy Depends on the Deployment and the Surface
Cody's documentation states that Sourcegraph collects prompts and responses to provide the service and collects usage data and feedback. It also states that this data is not used to train models. That is an important distinction, but it does not mean code and prompts never leave the developer's environment. Sourcegraph Enterprise offers single-tenant cloud and self-hosted options, and enterprise security controls apply across the platform. Buyers still need to map data flow for the chosen deployment, Cody client, Sourcegraph instance, and model provider.
A proper review should identify which repository excerpts Cody sends to an inference service, where conversations are stored, which administrators can access them, how retention is configured, which model providers are available, and whether prompts cross regions. Context filters can exclude selected repositories from chat and completions, but a filter is useful only if it matches the organization's data-classification policy and is tested. Permissions from connected code hosts must remain synchronized so an assistant never retrieves code a developer cannot normally see.
Sourcegraph's broader agent story adds another layer. MCP tools can expose code search, Deep Search, and other capabilities to external agents. This can reduce the waste of an agent blindly opening files, but it also creates a governed access path that needs authentication, scopes, tool controls, audit logs, and rate limits. Sourcegraph's August 2026 release added a general MCP tool blocklist, which is a practical control. Teams should start with the narrowest useful tool set and verify that repository permissions survive every hop from user to agent to Sourcegraph.
Context Quality Must Be Measured
Cody's promise depends on retrieval. If it selects the wrong files, the language model may produce a fluent answer grounded in irrelevant evidence. If an index is stale, an agent may reason about code that has already changed. If a monorepo contains multiple versions of a pattern, the nearest semantic match may not be the authoritative implementation. Better code search improves the odds of a good answer, but retrieval quality and task completion are separate measurements.
Sourcegraph itself now emphasizes this distinction. A useful enterprise evaluation should score whether the system retrieves the right files and symbols, whether the model interprets them correctly, and whether the resulting code or explanation solves the task. Those stages fail differently. Retrieval may omit a critical dependency. The model may ignore retrieved evidence. A correct answer may still violate a local convention or miss an operational constraint.
Build an evaluation set from real engineering work. Include navigation questions, unfamiliar-service explanations, bug investigations, cross-repository dependency checks, security searches, and small edits. Record the files and evidence a senior engineer expects the tool to find. Compare Cody, Deep Search, and an external agent using Sourcegraph MCP where relevant. Measure citation accuracy, missing evidence, time to a correct answer, model and credit cost, review effort, and the frequency of confident claims that cannot be verified.
Why the $16,000 Starting Price Changes the Audience
Sourcegraph's current enterprise plan starts at $16,000 and scales with team size. It includes credits for AI features, with organization-wide pooling and no monthly credit expiry according to the pricing page. It also includes the platform capabilities around search, Deep Search, Batch Changes, Insights, Monitoring, MCP, APIs, CLI access, enterprise administration, single-tenant cloud, and self-hosting. Volume pricing and additional credit buckets are available through sales.
The page does not publish a standalone Cody seat price. Buyers should confirm whether Cody is included in the proposed enterprise package, which users require licenses, how AI credits are consumed by Cody and other Sourcegraph features, how overages work, and what happens to unused credits at renewal. The $16,000 figure is a platform starting point, not enough information to forecast the final contract. Deployment, support upgrades, customer success, volume, and AI usage can change the total.
For a small team, this is difficult to justify if the goal is chat and completions. GitHub Copilot, Cursor, Tabnine, and other assistants offer clearer per-user entry points. Sourcegraph becomes financially plausible when the company has a big-code problem: many repositories, difficult onboarding, repeated migrations, security investigations that span systems, fragmented ownership, or agents that waste time and tokens searching for context. The return should be measured across these platform outcomes, not assigned entirely to Cody.
A Pilot for the Current Sourcegraph Platform
Begin with a codebase problem that individual assistants handle poorly. Choose several repositories with real dependencies and a task whose expected evidence is known. Connect the relevant code hosts, synchronize permissions, and confirm indexing freshness. Establish a baseline using the team's existing search and assistant workflow. Then test the same task through Cody in the IDE, Deep Search in Sourcegraph, and the Sourcegraph MCP server from the preferred external agent.
The interfaces should not be expected to produce identical outputs. Cody is suited to interactive development around the current work. Deep Search is designed for broader investigation and follow-up questions. MCP is useful when an implementation agent needs targeted code context without replacing its normal workflow. The pilot should reveal where each surface belongs, which reduces the temptation to deploy every feature to every developer.
Include a permissions test. Give different users access to different repositories and confirm that Cody, Deep Search, and MCP results respect those boundaries. Include stale or conflicting documentation. Rename a symbol or move an implementation and measure index lag. Ask a question with an incorrect premise and see whether the system challenges it. Run a cross-repository change only in an isolated test environment. Enterprise code intelligence earns trust when it handles incomplete and changing reality, not when it answers a carefully prepared demo question.
The pilot also needs human outcomes. Does a new engineer reach the right service owner faster? Does a staff engineer spend less time answering navigation questions? Can a security team prove the scope of a vulnerable pattern? Can a migration owner identify every affected repository and produce reviewable changes? Do coding agents make fewer exploratory tool calls when Sourcegraph provides context? These are more meaningful than counting generated lines or chat messages.
Where Cody Fits in 2026
Cody is still useful for enterprise developers who want chat, completion, edits, and prompts inside their IDE with Sourcegraph's repository-scale context. It is especially relevant to organizations that already use Sourcegraph or are buying the wider platform for code understanding. The supported clients and ongoing model updates show that Cody Enterprise remains a live product, even as Deep Search has taken over the main web experience.
Cody is not a sensible recommendation for someone searching for a free AI coding extension. Those plans ended more than a year ago. It is also not the centre of Sourcegraph's public identity now. Sourcegraph increasingly supplies intelligence to humans and many different agents, while Amp operates independently as an agentic coding company. Keeping those facts clear protects users from paying attention to an obsolete comparison.
The WhatAI Verdict
Sourcegraph Cody has not disappeared, but the product around it has changed enough that its old directory profile must be retired. Cody Enterprise remains supported, receives model updates, and continues to offer code-aware assistance in major IDEs. Individual Cody Free and Pro plans no longer exist. The current purchase is Sourcegraph Enterprise, starting at $16,000, with Cody sitting alongside Deep Search, code search, MCP context, Batch Changes, and governance for complex codebases.
That makes Cody narrower and Sourcegraph more interesting. The strongest reason to buy is not autocomplete. It is the possibility of giving developers and coding agents a shared, permission-aware understanding of a sprawling code estate, then using that understanding for investigation, navigation, security work, and coordinated change. Organizations with genuine big-code problems should evaluate the platform on their own repositories. Everyone else should treat old Cody Free and Pro recommendations as expired and choose a current individual coding assistant instead.
Sourcegraph Cody remains a supported enterprise AI coding assistant for chat, completions, edits, prompts, and repository-aware development. Its former Free and Pro plans ended in July 2025, so current access belongs within Sourcegraph Enterprise rather than a consumer subscription.
Status, Pricing, Context, and Alternatives
Sourcegraph Enterprise starts at $16,000 and combines code understanding, Deep Search, Code Search, MCP, Batch Changes, Insights, Monitoring, private deployment, and AI credits. Cody is the IDE assistance layer inside that wider platform.
Should Your Team Still Use Cody?
Cody fits enterprises with large or multi-repository codebases that need IDE assistance grounded in Sourcegraph context. Individuals seeking a free or low-cost coding assistant should choose a current alternative because Cody Free and Cody Pro no longer exist.
About Sourcegraph Cody
Sourcegraph Cody is an enterprise AI coding assistant that combines language models with repository context from Sourcegraph's code search and intelligence platform. It supports chat, code completions, edits, custom prompts, context filters, and model selection in Visual Studio Code, JetBrains IDEs, Visual Studio, and Sourcegraph's web environment. Cody Free and Cody Pro were discontinued in July 2025, while Cody Enterprise remains supported and actively updated. It now sits within Sourcegraph's broader enterprise platform alongside Deep Search, Code Search, MCP, Batch Changes, Insights, and Monitoring.
Use Cases
Key Features
- ✓ Developer chat grounded in local and remote code
- ✓ Current-line and multi-line code completions
- ✓ Code edits and customizable prompts
- ✓ Repository, symbol, API, and usage context
- ✓ Context filters for selected repositories
- ✓ Multiple supported LLMs
- ✓ VS Code, JetBrains, Visual Studio, and web clients
- ✓ Integration with Sourcegraph Code Search
- ✓ Single-tenant cloud and self-hosted platform options
- ✓ Enterprise administration and security controls
- ✓ Shared platform credits for AI features
Pricing
Cody Free
Discontinued
- • Ended July 23, 2025
- • No new individual access
Cody Pro
Discontinued
- • Ended July 23, 2025
- • Former users were directed to Amp
Sourcegraph Enterprise
Starting at $16,000
- • Cody Enterprise support
- • Credits for AI features
- • Deep Search, Code Search, and MCP
- • Single-tenant cloud or self-hosted
- • Volume pricing through sales
Pricing varies by plan and region — see current pricing.
Plan features change — last updated: 2026-08-24.
Details
Tags
Sourcegraph Cody Community Discussions
Explore community discussions. Ask and answer questions on Sourcegraph Cody to grow and learn together.
Sourcegraph Cody Showcase
Sourcegraph Cody — Frequently Asked Questions
Is Sourcegraph Cody discontinued?
Cody Free and Cody Pro were discontinued on July 23, 2025. Cody Enterprise remains supported and actively updated within Sourcegraph Enterprise.
How much does Sourcegraph Cody cost?
Sourcegraph no longer publishes an individual Cody price. The Sourcegraph Enterprise platform starts at $16,000 and includes AI credits. Buyers should confirm Cody licensing, users, credits, and final contract cost with sales.
What replaced Cody Free and Cody Pro?
Sourcegraph directed former individual Cody users to Amp. Sourcegraph and Amp became independent companies in December 2025, so Amp is not a current Cody plan.
Which IDEs support Cody Enterprise?
Current Sourcegraph documentation lists Visual Studio Code, JetBrains IDEs, Visual Studio, and the Sourcegraph web application.
What does Cody use for codebase context?
Cody uses Sourcegraph's Search API to retrieve context from local and remote repositories, including files, symbols, APIs, definitions, and usage patterns.
Does Cody train models on customer code?
Sourcegraph states that it collects prompts, responses, usage data, and feedback to provide and improve the experience, but does not use customer data to train models. Exact storage and data flow depend on deployment and providers.
What is the difference between Cody and Deep Search?
Cody is an IDE-oriented coding assistant for chat, completions, and edits. Deep Search is Sourcegraph's broader AI investigation surface and replaces Cody Web when enabled in Sourcegraph 7.
Can Sourcegraph context work with other agents?
Yes. Sourcegraph's MCP server can provide code search and context to tools such as Claude Code, Cursor, Codex, and Amp, subject to enterprise configuration and permissions.
Sources & References
Try Sourcegraph Cody
Visit the official website to get started with Sourcegraph Cody today.
Visit Sourcegraph Cody →