# Recce

> AI data review agent for dbt pull requests, with open-source data diff, schema diff, downstream impact analysis, and column-level lineage.

Source: https://toolsfordata.com/tools/recce/ · Updated 2026-11-13

[Website](https://reccehq.com/) · [Docs](https://docs.reccehq.com/) · [Pricing](https://reccehq.com/pricing/) · [GitHub](https://github.com/datarecce/recce) (480+ stars)

Best for: dbt teams that want reviewers to see the data impact of code changes before approving a pull request.

## Overview

- Core Strength: dbt pull request review — data diffs, column lineage and impact radius before merge
- Test Types: Partial — table compare, data diff, failed rows
- Detection: Partial — profiling
- Monitor Coverage: 0 of 11 types
- Incident Workflow: Not supported
- Quality Analytics: Not supported
- Lineage Depth: Partial — table, column, impact analysis
- Governance Overlap: Not supported
- FinOps & AI Observability: Not supported
- AI Agent Features: Not confirmed
- Developer Surface: Partial — MCP, dbt, CI/CD
- Deployment Model: Partial — SaaS, self-hosted
- Security & Access: Partial — SSO, RBAC, certifications
- Pricing: From $250/month (1,000 reviews)
- Free Option: Open source (Recce OSS) · Free tier (100 reviews/month)
- Notable Caveat: No monitors, alerts or schedules; PR gating and AI summary are Cloud-only

## What it's used for

Recce is used to review dbt pull requests with data, not just code: it diffs your branch against production and shows which rows, columns and downstream models change before you merge.

## Pricing

Recce's open-source version is free under Apache 2.0. Recce Cloud has a free plan with 100 reviews per month, a $250 per month plan with 1,000 reviews, and custom enterprise pricing.

## Pricing

- Model: per agent review
- Free: Free (100 reviews/month)
- Open source: OSS CLI (Recce OSS)
- Entry team price: $250/month (1,000 reviews)
- Entry enterprise price: Custom
- Source: https://reccehq.com/pricing/ (checked 2026-11-02)

## Features

### Data Testing

- SQL Table Tests: No — No pass/fail rules; custom SQL runs as Query Diff, which compares results between base and PR branch. ([source](https://docs.reccehq.com/technical-concepts/configuration.md))
- DataFrame Tests: No — Recce queries warehouse tables through dbt adapters, not DataFrames. ([source](https://docs.reccehq.com/setup-guides/connect-to-warehouse.md))
- Unstructured Data Tests: No — No LLM or JSON Schema checks. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Table Comparison Tests: Yes — Row Count Diff and Profile Diff compare row counts and column statistics between the production and PR versions of a dbt model. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Data Diff: Yes — Value Diff joins base and PR versions on a primary key and reports added, removed and changed rows with a match % per column. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Data Contracts: No — Schema Diff shows added, removed and changed columns, but there are no contracts. ([source](https://docs.reccehq.com/what-you-can-explore/lineage-diff.md))
- Inspect Failed Rows: Yes — Value Diff lists the rows whose values changed, were added or removed. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))

### Data Observability

- Data Profiling: Yes — Profile Diff shows row count, not-null %, distinct count, min, max and average per column, before and after the change. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Anomaly Detection: No — No learned thresholds; diffs compare two versions of the data. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Freshness Monitors: No — No freshness checks; Recce reviews changes, it doesn't monitor production. ([source](https://docs.reccehq.com/collaboration/preset-checks.md))
- Volume Monitors: No — Row Count Diff compares counts on a pull request, not over time. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Schema Monitors: No — Schema Diff compares columns on a pull request; there are no alerts on production schema changes. ([source](https://docs.reccehq.com/what-you-can-explore/lineage-diff.md))
- Field Monitors: No — Profile Diff is a one-off comparison, not a time series. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Distribution Monitors: No — Top-K and Histogram Diff compare distributions on a pull request, not over time. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Segment Monitors: No — No group-by monitors. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Custom SQL Monitors: No — Query Diff runs custom SQL on a pull request; nothing is tracked over time. ([source](https://docs.reccehq.com/technical-concepts/configuration.md))
- Partition-Based Monitors: No — No partition or time filters on checks. ([source](https://docs.reccehq.com/technical-concepts/configuration.md))
- Window-Based Monitors: No — No time-window metrics. ([source](https://docs.reccehq.com/technical-concepts/configuration.md))
- UI-Based Monitors: No — Preset checks can be set up in Recce Cloud, but they run on pull requests, not as monitors. ([source](https://docs.reccehq.com/collaboration/preset-checks.md))
- Monitors-as-Code: No — recce.yml defines preset diffs that run on each pull request, not monitors. ([source](https://docs.reccehq.com/technical-concepts/configuration.md))
- Pipeline Monitoring: No — No pipeline run monitoring. ([source](https://docs.reccehq.com/whats-recce/cloud-vs-oss.md))

### Incident Management

- Alert Notifications: No — Results appear as pull request comments and status checks; no email or Slack alerts. ([source](https://docs.reccehq.com/using-recce/data-reviewer.md))
- Incident Triage: No — No incidents; reviewers approve checks on a pull request. ([source](https://docs.reccehq.com/using-recce/data-reviewer.md))
- Alert Feedback: No — No alerts to give feedback on. ([source](https://docs.reccehq.com/collaboration/checklist.md))
- Incident Resolution: No — No incident workflow; checks are approved or not. ([source](https://docs.reccehq.com/collaboration/checklist.md))
- Ticketing Integrations: No — Integrations are GitHub, GitLab and warehouses only. ([source](https://docs.reccehq.com/setup-guides/connect-git.md))
- On-Call Integrations: No — No PagerDuty or Opsgenie. ([source](https://docs.reccehq.com/setup-guides/connect-git.md))

### Data Quality Analytics

- Data Quality Dashboards: No — No data quality dashboard; Enterprise lists an analytics dashboard on review activity. ([source](https://reccehq.com/pricing/))
- Data Health Scores: No — No score per table; each check is pass, warning or fail. ([source](https://docs.reccehq.com/using-recce/data-reviewer.md))
- Incident Resolution Metrics: No — No incident metrics. ([source](https://docs.reccehq.com/using-recce/data-reviewer.md))

### Data Lineage

- Table-Level Lineage: Yes — Lineage Diff shows added, modified and removed dbt models and their downstream models. ([source](https://docs.reccehq.com/what-you-can-explore/lineage-diff.md))
- Column-Level Lineage: Yes — Column lineage traces a column through dbt models by parsing their SQL. ([source](https://docs.reccehq.com/what-you-can-explore/column-level-lineage.md))
- End-to-End Lineage: No — Lineage covers dbt models only, not ingestion or BI tools. ([source](https://docs.reccehq.com/what-you-can-explore/lineage-diff.md))
- Impact Analysis: Yes — Impact Radius lists the downstream models and columns a change affects. ([source](https://docs.reccehq.com/what-you-can-explore/impact-radius.md))

### Data Governance

- Data Catalog: No — No asset catalog; views cover the models in one dbt project. ([source](https://docs.reccehq.com/whats-recce/cloud-vs-oss.md))
- Data Catalog Integration: No — No push of results to catalogs. ([source](https://docs.reccehq.com/setup-guides/connect-to-warehouse.md))
- Data Classification: No — No PII detection. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Data Products: No — Projects map to repositories, not data products. ([source](https://docs.reccehq.com/using-recce/admin-setup.md))

### FinOps

- Query Performance: No — Diffs show a rough cost label, but no query performance metrics. ([source](https://docs.reccehq.com/what-you-can-explore/data-diffing.md))
- Cost Monitoring: No — No warehouse cost tracking. ([source](https://docs.reccehq.com/whats-recce/cloud-vs-oss.md))

### LLMOps

- AI Observability: No — No monitoring of LLM or agent outputs. ([source](https://docs.reccehq.com/what-you-can-explore/summary.md))

### Workflows

- Scheduling: No — Checks run on pull requests or on demand; there is no schedule. ([source](https://docs.reccehq.com/collaboration/preset-checks.md))
- CI/CD Integrations: Yes — Runs in GitHub Actions and GitLab CI on each pull request; Recce Cloud adds PR comments, status checks and merge blocking. ([source](https://docs.reccehq.com/setup-guides/setup-ci.md))
- Circuit Breakers: No — Recce reviews pull requests by diffing dev against production; it runs before merge, not in production pipelines. ([source](https://docs.reccehq.com/))
- Webhooks: No — No outbound webhooks. ([source](https://docs.reccehq.com/technical-concepts/configuration.md))
- REST / GraphQL API: Unknown — No public API is documented. ([source](https://reccehq.com/security/))
- MCP Server: Yes — recce mcp-server gives Claude Code, Cursor and other agents lineage, schema, row count, value, profile and query diff tools. ([source](https://docs.reccehq.com/setup-guides/mcp-server.md))
- Python SDK: No — Only the recce and recce-cloud command-line tools are documented. ([source](https://docs.reccehq.com/using-recce/cli-commands.md))
- dbt Integration: Yes — Built for dbt: reads manifest.json and catalog.json from the production and PR runs of a dbt project. ([source](https://docs.reccehq.com/what-you-can-explore/lineage-diff.md))

### AI Agents

- AI Chat Assistant: Unknown — The website says users can ask the agent follow-up questions, but no in-product chat is documented. ([source](https://reccehq.com/))
- Natural Language Test Generation: No — Agents can create checks through the MCP server, but Recce itself has no plain-English check builder. ([source](https://docs.reccehq.com/setup-guides/mcp-server.md))
- Automated Test Generation: No — The Recce Cloud agent picks which diffs to run on a pull request, but does not create lasting checks from profiling. ([source](https://docs.reccehq.com/what-you-can-explore/summary.md))
- AI Issue Triaging: No — The Recce Cloud agent summarizes what a pull request changes in the data; it does not diagnose data quality failures. ([source](https://docs.reccehq.com/what-you-can-explore/summary.md))
- AI Issue Resolution: No — The summary suggests next steps, but no fix is generated. ([source](https://docs.reccehq.com/what-you-can-explore/summary.md))

### Security

- SSO: Yes — Recce Cloud supports SSO on paid plans. ([source](https://docs.reccehq.com/using-recce/admin-setup.md))
- SCIM Provisioning: Unknown — Members are invited by email; SCIM is not mentioned. ([source](https://docs.reccehq.com/using-recce/admin-setup.md))
- RBAC: Yes — Recce Cloud has Owner, Admin and Member roles, with project-level permissions on Enterprise. ([source](https://docs.reccehq.com/using-recce/admin-setup.md))
- Audit Log: Unknown — Recce Cloud logs activity per check, but no organization-wide audit log is documented. ([source](https://docs.reccehq.com/collaboration/activity.md))
- Compliance Certifications: Yes — Recce shows a SOC 2 badge linking to its trust center. ([source](https://reccehq.com/))

### Deployment

- SaaS Deployment: Yes — Recce Cloud is a hosted service connected to your Git provider and warehouse. ([source](https://docs.reccehq.com/whats-recce/cloud-vs-oss.md))
- Hybrid Deployment: Unknown — Enterprise lists Bring Your Own Cloud, but a split control plane and data plane is not described. ([source](https://reccehq.com/pricing/))
- Self-Hosted Deployment: Yes — The open-source CLI runs on your own machine or CI, and Enterprise can be deployed in your VPC or on-premises. ([source](https://reccehq.com/pricing/))
- Cloud-Native App: No — Runs as a local CLI or as Recce Cloud, not inside the warehouse. ([source](https://docs.reccehq.com/setup-guides/connect-to-warehouse.md))

### Open Source

- Open Source: Yes — Recce's core is open source under Apache 2.0; Recce Cloud is the paid version. ([source](https://github.com/DataRecce/recce))

## Integrations

8 integrations, each sourced from the vendor docs.

- Code Repositories: GitHub, GitLab
- Data Orchestration: GitHub Workflows
- Data Transformation: dbt
- Data Warehouse: Amazon Redshift, Databricks SQL, Google BigQuery, Snowflake

## Alternatives

- [Datafold](https://toolsfordata.com/tools/datafold/): The commercial data-diff pioneer, now focused on migrations: its pull-request app is closed to new customers.
- [DataComPy](https://toolsfordata.com/tools/datacompy/): Consider DataComPy for free table and DataFrame diffs outside dbt, in a notebook or CI.
- [diffly](https://toolsfordata.com/tools/diffly/): Consider diffly for fast Polars diffs with per-column match rates, inside tests.
- [dbt tests](https://toolsfordata.com/tools/dbt-tests/): Pair it with dbt tests: they check rules, Recce shows what changed.
- [Elementary OSS](https://toolsfordata.com/tools/elementary-oss/): Add Elementary OSS to monitor the same dbt project after merge.
- [Foundational](https://toolsfordata.com/tools/foundational/): Consider Foundational for pull-request impact analysis beyond dbt, from source code to BI.
