Deep Search can now use a dedicated subagent specialized for finding files relevant to your query in case a specific code domain needs to be researched. This subagent performs scoped searches and presents a summary to the main agent. Since the main Deep Search agent only needs to read this summary, this saves tokens in its context window.

This means Deep Search can explore more of your codebase while using fewer tokens, so you get more focused context and longer conversation. We expect this feature will significantly alleviate issues with deep single-question research - for multi-question, we just shipped auto-compaction.

We are removing the need to manually manage your context window limits!

Compaction runs in the background, so there is nothing new to configure or learn. Keep asking follow-up questions, and Deep Search will make room when older conversation history starts to crowd the context window.

Deep Search conversation history now has a Shared with me tab. It lists conversations you opened from a shared Deep Search link, separate from conversations you created yourself.

From the shared list, you can reopen, export, fork, or remove a conversation from your shared history without deleting the owner's original conversation.

Administration on Sourcegraph has historically been an all-or-nothing decision: either a user was an administrator and could do everything, or they weren't and couldn't access any admin surface. We've replaced this with a set of fine-grained RBAC permissions that map onto distinct admin responsibilities. Now you can grant users exactly the access they need and nothing more.

Eight new permission namespaces are available in the role editor (Site admin → Users & auth → Roles), each grantable independently to any role.

We are introducing a new compare page for viewing changes between two revisions.

Deep Search has always been good at finding and explaining code. It can now also count, rank, and aggregate it without blowing up the context window.

A new sandboxed scripting tool lets the agent run programmatic aggregations across many searches in a single turn. Questions that don't fit into a single query, e.g. top-N rankings, cross-search joins, distributions across many files, can now be answered end-to-end, with every underlying search still cited so you can verify how the answer was assembled.

When precise code intelligence is available, you now get a summary of the meaning of the symbol, and usage of the symbol across the codebase. In combination with the precise code intelligence hover information, this often saves the need to jump to a symbol and back which interrupts the flow of reading, and simplifies understanding complex codebases. Linked citations can even serve as an alternative to traditional code navigation.

Search contexts are used to scope searches in code search. Now, you can do the same for Deep Search to guide the response to a specific domain. The users default search context is pre-selected.

Deep Search can now search Git tags, ancestry, and commits on any branch to pinpoint where a change lives in your Git history.

For example, ask Deep Search: "In Kubernetes v1.29, a regression was introduced where the apiserver_watch_events_sizes metric stopped reporting total outgoing watch event traffic. How many versions shipped with the bug before it was resolved?"

You can now rename your Deep Search conversations to keep your workspace organized! Click directly on the conversation title to edit it inline, or use the rename option from the conversation's options menu.

While syncing individual changesets was already possible, Batch Changes now supports syncing changesets as a bulk action. For batch changes with hundreds or thousands of changesets, this means you can refresh the state of all changesets from the code host in a single click—without having to load and sync each one individually.

This is especially useful when changesets have been updated externally (e.g., merged or closed on the code host) and you want Sourcegraph to reflect the latest state immediately across the entire batch change.

Batch Changes now supports OAuth-based authentication for GitHub, joining the existing support for GitLab, Bitbucket Server, and Bitbucket Cloud. Users can now authorize Batch Changes to act on their behalf using OAuth instead of manually creating and pasting personal access tokens.

When adding a credential for a GitHub code host, users will see a new OAuth option in the authentication strategy dropdown. Clicking through initiates a standard OAuth flow that automatically provisions the credential with the correct scopes.

In Code Search, you can now describe what you're looking for in plain English — press # in the search bar to activate query assist, type something like TODO comments in python files or what did alice work on last week?, and our custom-trained model will translate it into a precise Sourcegraph search query.

Use Deep Search to create architecture diagrams, technical documentation, or research summaries, then export and share them with your team.

As previously announced, the default Sourcegraph MCP endpoint at /.api/mcp now exposes a smaller, more focused set of tools for common context-gathering workflows.

This reduces tool-list noise for MCP clients and avoids spending context window budget on lower-frequency tools that most agents do not need by default.

Batch Changes now supports OAuth-based authentication for Bitbucket Server and Bitbucket Cloud, joining the existing support for GitHub and GitLab. Users can now authorize Batch Changes to act on their behalf using OAuth instead of manually creating and pasting personal access tokens.

When adding a credential for a Bitbucket Server or Bitbucket Cloud code host, users will see a new OAuth option in the authentication strategy dropdown. Clicking through initiates a standard OAuth flow that automatically provisions the credential with the correct scopes.

If your organization makes use of multiple Slack workspaces, or Slack Enterprise Grid, you may want Sourcegraph to act as the central code understanding platform where anyone across your organizations can use Sourcegraph to understand your entire codebase.

The Sourcegraph Slack integration now supports connecting any number of Slack workspaces, including Enterprise Grid workspaces, to the same Sourcegraph instance - enabling your Slack users to ask @Sourcegraph questions about your codebase directly in their Slack workspace and get Deep Search answers.

Hit the context window limit, or want to explore a different direction without losing your original thread? Use the Fork button to start a new conversation that picks up where the original left off.

You can now log in to src with OAuth, making it faster to connect src to your Sourcegraph instance without manually managing an access token.

We’ve also added src auth token, which prints the token for your current authenticated session. That makes it easier to use src in scripts and shell workflows, for example by passing $(src auth token) to other commands when needed.

Paste a Deep Search URL or read token into your message, and the agent will load that investigation as context for your new prompt. For example:

• Read https://sourcegraph.com/deepsearch/b875dd16-a58f-4c27-84b8-4b4306e49df7 and tell me what else we should investigate about our flaky tests
• Based on the findings in https://sourcegraph.com/deepsearch/shared/e3c0150a-b7ef-4955-bed4-d5820ca7a70d, investigate whether the same logging patterns appear in our other microservices
• What did we learn about authentication in https://sourcegraph.com/deepsearch/71a48f28-d241-4efa-bc62-4afde5499b00? Are there similar issues in the authorization code?
• Continue the investigation from https://sourcegraph.com/deepsearch/7347ddaa-d579-4b91-b152-ca27e4ddb656 by checking if those dead code patterns exist in our newer repositories
• Compare what we found in https://sourcegraph.com/deepsearch/e782c115-2dfa-44a5-8a88-2fad218a2fe1 with the current state of the codebase - has the performance issue been addressed?
• Look at https://sourcegraph.com/deepsearch/shared/3f1beb2b-bded-4fda-96cc-1af7192f24b6 and investigate whether those security concerns apply to our newer API endpoints

We've improved the set of tools available for agents inside Sourcegraph to emit more token-efficient outputs. This includes tools that Deep Search uses under the hood, as well as tools available in our MCP server.

Instead of JSON, the tools now emit a more compact plaintext format, saving tokens on every execution. A lower token count also leads to lower latency and costs. The new format is also more selective about which pieces of information it shows to the agent.

Sourcegraph Analytics now includes pacing charts for Deep Search, making it easy for admins to track usage against their purchased quota as the month progresses. The charts show both burndown and cumulative views for the current month, mapped against the previous month so you can spot trends at a glance.

The burndown view tracks usage against your purchased amount, helping you know when it’s time to top up.