What's New in Recce

Review PR Data Impact Directly from Your Editor

Running data diffs meant switching between Recce and your editor, copying model names, and pasting results back. With eight new MCP tools, AI agents can now access every evidence-gathering feature directly.

The new tools include value_diff for row-level comparisons, top_k_diff for categorical distributions, histogram_diff for numeric distributions, and get_cll for column-level lineage. Combined with the existing tools, agents now have 16 MCP tools to review data impact without leaving your workflow.

See details here

---

AI Summary Without a Warehouse Connection

Users without a connected warehouse could not get AI-generated summaries of their PR impact. The summary feature required live data access, which blocked adoption for teams still setting up their environment.

Recce now generates AI summaries based on available metadata even when no warehouse is connected. Teams evaluating Recce Cloud see value immediately, and the summary explains what additional insights a warehouse connection would unlock.

---

Agent Asks for Help When Stuck

During onboarding, the web agent sometimes hit edge cases it could not resolve on its own. Users had no clear path to get help, and the team had no signal about which scenarios needed improvement.

The agent now detects when it cannot make progress and offers to escalate. Users get a direct path to support, and the team collects structured feedback on which onboarding scenarios need attention.

---

Python 3.9 Deprecation

Python 3.9 reached end-of-life in October 2025, and Recce will drop support in a future release. Upgrade to Python 3.10 or later to avoid interruptions.

Recce Cloud Instances Launch in Seconds

Launching a Recce Cloud instance used to take 30 seconds or more. Every PR review started with a wait while a fresh container spun up.

Recce Cloud now draws instances from a pre-warmed container pool. Startup drops from 30 seconds to under 7 seconds. The pool stays warm so the next launch is just as fast. Available on the Enterprise plan.

---

MCP Server: Stable Setup with Clear Error Messages

The Recce MCP server crashed without a useful error when base artifacts were missing. Users had to dig through logs to find what went wrong. The --help text did not mention prerequisites.

The MCP server now classifies errors and tells you exactly what is missing. A new single-env onboarding mode works without a full two-environment setup. The --help output documents prerequisites so AI agents and users know what Recce MCP needs before starting.

---

Sort Value Diff Columns by Match Percentage

Value Diff showed columns in their original order. Finding the columns with the biggest differences meant scanning every row and comparing percentages manually.

Columns now sort by Matched % by default, with the lowest match at the top. Click any column header to change the sort order. The columns that need attention surface immediately.

---

See Which Columns Changed at the SQL Level in Schema View

Schema View showed added and removed columns, but columns with modified SQL definitions looked the same as unchanged ones. You had to open the code to check whether a column’s logic actually changed.

Schema View now marks columns with ~ (modified), + (added), and - (removed) badges. Columns where the underlying SQL changed show a definition-changed badge. You can spot logic changes without leaving the schema.

---

Column Lineage Resolves Recursive CTEs

Recursive CTEs — common in models that traverse hierarchies or generate sequences — showed “unknown” in column lineage. The self-referential nature of these CTEs broke lineage tracing.

Column lineage now follows recursive CTE references through each iteration. Models that use WITH RECURSIVE display correct lineage paths instead of dead ends.

---

Bug Fixes

Lineage view stays focused during row count. Running a row count from NodeView no longer resets the zoom and pan position.

Result pane appears in the correct position. The result pane no longer jumps to the top of the page.

Source nodes display their schema. Source nodes with columns in the catalog now show those columns instead of an empty panel.

AI-Assisted Onboarding

Setting up Recce involves connecting git, configuring your data warehouse, and adding Recce to CI/CD. The first two steps are straightforward with UI guidance. CI/CD setup is where it gets complex—every team has different pipelines, platforms, and configurations.

Recce onboarding now includes an AI assistant that handles the hardest part. When you reach CI/CD setup, the agent generates workflow configurations for your specific setup—GitHub Actions, GitLab CI, or other platforms. Describe your pipeline and the agent produces the YAML you need.

---

See Changes and Impact at a Glance

Finding what changed in large diffs required scanning through every line. Click the model to view all columns without highlighting changes. Value distributions lacked context for quick comparison.

Recce now makes changes and their impact immediately visible:

• Schema change indicators — Visual badges mark columns as added, removed, or modified
• Collapsible unchanged regions — Unchanged code collapses automatically so you focus on what matters
• TopK inline labels — Distribution bars display numeric values inline for quick comparison

---

Customize Lineage Graph Node Shapes

Lineage graphs used the same node shape for all resource types. Distinguishing between models, sources, and seeds required reading labels.

You can now configure custom node shapes in Mermaid lineage exports. Different shapes for different resource types make the graph easier to scan. Thanks to Daisuke Taniwaki for this community contribution (PR#1137).

---

See Instance Status on Your Project Page

Opening a project meant waiting without knowing if the instance was already running or needed a cold start. There was no way to anticipate the wait time before clicking in.

The project page now shows instance status indicators. Know whether an instance is already running or requires a cold start before you click in.

---

Export Query Results to Excel and TSV

Copying data diff results for analysis outside Recce required manual formatting. You had to copy raw data and reformat it in your spreadsheet tool.

Recce now offers three export options for data diff results: copy as TSV to clipboard, download as Excel, or download as TSV file. Select your preferred format from the export menu and paste or open directly in your spreadsheet tool.

---

One Account for GitHub and Email Login

Users who signed up with both GitHub and email had two separate Recce Cloud accounts. Switching between them meant losing access to projects tied to the other account.

Recce Cloud now merges accounts that share the same email address. The next time you log in with GitHub, your GitHub and email accounts automatically combine into one. All your projects and settings remain accessible.

Web Agent Configures Your PR Data Review

Setting up Recce means configuring CI/CD to trigger data review on every PR. Every team’s CI/CD setup is different—custom runners, matrix builds, environment variables—and getting Recce to fit takes trial and error.

The new Web Agent reads your existing workflow files and generates the right configuration for you. Connect git, let the agent handle CI/CD, link your data warehouse. Three steps to automated data review on every PR.

GitHub Actions supported today, with more CI platforms coming.

Try it at cloud.reccehq.com/onboarding/get-started

This release makes Lineage Diff capture exactly what you see, plus early access to AI-assisted setup.

Lineage Diff: Add Exactly What You See

Adding Lineage Diff checks is simpler. When viewing your lineage filtered by selectors, with Column-Level Lineage or Impact Radius applied, you can add that exact view to your checklist with one click.

Now what you see is what you add. Your current lineage view settings carry through to the check.

Incremental Model Awareness in Impact Analysis

Recce Cloud now recognizes incremental models and reflects this in impact analysis. When you modify a model with incremental logic, the data review shows accurate scope.

Example: A PR that adds a date filter to orders_daily_summary correctly shows the row count change from 730 to 364 rows, with clear indication that the change is isolated with no downstream impact

Experimental: Recce Plugin for Claude Code

We’re exploring AI-assisted onboarding for Recce Cloud. Set up CI/CD configuration by chatting with Claude.

Note: This is experimental. The onboarding flow is incomplete (Data Warehouse and Git setup not yet supported). We’re sharing early to signal our direction and gather feedback.

Interested? Install https://github.com/DataRecce/recce-claude-plugin

📂 Monorepo Support

Recce now supports monorepo setups. Teams with multiple data projects within a single repository.

👍 Data Review Feedback

Rate data review with 👍 or 👎. Your feedback help our agent to generate better summaries over time.

⚡ Faster Instance Launch

Recce instances now launch in ~30 seconds, down from 60 seconds. Less waiting, more reviewing.

🎨 Unified UI Components

Same polished components across Recce OSS and Cloud. Dark/light mode, consistent styling, better quality everywhere.

🏠 Updated Homepage

Our homepage https://reccehq.com now reflects Recce’s evolution as an AI Data Review Agent with refreshed visuals and messaging.

🤖 Smarter Data Agent Review

Recce Data Agent review now consume checklist results, giving you richer context in every review. Summaries also trigger automatically when you upload artifacts, no manual steps needed.

🔗 Instant team collaboration

Share a link and teammates join the same PR review instantly, no setup needed.

📥 CSV Download

Export query results to CSV directly from Recce.

Diff Options at a Glance

Available diff types are now displayed at the top of the model panel. See your options clearly without opening a dropdown menu.

🔐 Databricks OAuth Authentication

Connect to Databricks using OAuth client-based authentication for more secure, token-free access.

🌙 Dark Mode

Recce now supports full dark mode across the entire interface. Your preference is automatically saved, so Recce remembers your choice between sessions.

🎨 Clearer Data Comparisons

Base and current values now display side-by-side with delta percentages, making it easier to understand the magnitude of changes at a glance.

💬 Checklist Comments Sync to PR

Comments added to your validation checklist now automatically post to your GitHub PR. Keep reviewers informed without switching contexts or duplicating your notes, everything stays in sync.

🔗 Customize Your Organization Slug

You can now change your organization’s URL slug directly from settings. Perfect for teams transitioning from a personal account to a proper organization identity.

✨ Improvements

• Profile Query Result Table: Improved presentation for easier reading
• Compressed Sidebar Columns: Base and current columns for Type and Column ID are now collapsed, giving you a cleaner, more efficient view

• Frontend Infrastructure: Continued upgrades to keep Recce’s interface modern and performant

AI Summary for Dev Sessions

Generate AI summaries for your local development work before opening a PR. Upload your dev session to Recce Cloud and get an AI-powered overview of your changes, helpful for self-review or sharing context with teammates early.

Improved AI Summary for Large Lineage

When your lineage includes more than 30 models, the AI summary now generates a cleaner, more readable mermaid diagram. No more unreadable graphs for complex projects.