Loading tool details…
Databricks Mosaic AI Agent Framework (marketed as Agent Bricks in 2026) is Databricks' end-to-end platform for building, evaluating, and serving production AI agents on top of the Lakehouse — with unified governance, native RAG over Unity Catalog data, and consumption-based pricing in Databricks Units (DBUs).
Databricks Mosaic AI Agent Framework (marketed as Agent Bricks in 2026) is Databricks' end-to-end platform for building, evaluating, and serving production AI agents on top of the Lakehouse — with unified governance, native RAG over Unity Catalog data, and consumption-based pricing in Databricks Units (DBUs).
Mosaic AI Agent Framework is Databricks' answer for teams that want their AI agents to live where their data already lives. The framework spans the full lifecycle: Vector Search (Databricks' managed vector index built on Delta tables), Model Serving (endpoints for Meta Llama, Anthropic Claude, and Databricks' own DBRX models), the Agent Framework SDK (a Python-first API modeled after LangGraph/LangChain patterns), MLflow for tracing and evaluation, and Databricks Apps for serving the agent as a Lakehouse-native web app. Databricks positions this under the newer 'Agent Bricks' brand alongside a governance layer that piggybacks on Unity Catalog for row-level access control.
Pricing is consumption-based via Databricks Units (DBUs). You are billed for the underlying compute — Serverless Model Serving DBUs for LLM inference, Vector Search DBUs for indexing and retrieval, and Serverless Compute DBUs for agent runtime — on top of your cloud provider's infrastructure (AWS/Azure/GCP) charges. Databricks does not publish per-token retail prices for third-party models on the marketing page; costs are visible in the cluster/serving cost estimator once you're inside the workspace. This makes accurate budgeting harder than a flat SaaS but scales down to zero when idle on serverless endpoints.
Mosaic AI is the correct choice when your data of record already lives in Databricks and unified governance across data + models + agents is more valuable than picking the absolute best point tool for each layer. If your data lives outside the Lakehouse, alternatives like /tools/langgraph on /tools/e2b or a native /tools/anthropic + /tools/pinecone stack usually integrate faster.
Revolutionary approach that eliminates manual trial-and-error agent development through Instructed Retriever technology, which automatically learns optimal retrieval strategies for each domain and query pattern, improving relevance by 15–25% over standard vector-search RAG.
Pre-built agent architectures optimized for common enterprise scenarios: Information Extraction agents for structured data extraction from documents, Knowledge Assistants for Q&A over document corpora, SQL Agents for natural-language analytics, and custom agents for specialized workflows.
Deep integration with Unity Catalog that enables agents to understand enterprise context including table schemas, column descriptions, data lineage, and access policies — allowing agents to answer questions with full awareness of organizational data assets.
Access to leading AI models from OpenAI, Anthropic, Google, Meta, and open source through the AI Gateway, with intelligent routing, cost tracking, rate limiting, and guardrails applied consistently across all model providers.
Comprehensive platform for monitoring, tracing, and optimizing AI agents with integrated experiment tracking, automated evaluation datasets, LLM-as-a-judge scoring, and production quality dashboards for continuous improvement.
Advanced capability that automatically creates domain-specific synthetic data resembling production queries, enabling teams to build robust evaluation suites and stress-test agents before deployment without relying solely on manually curated test sets.
Consumption-based
Free
Custom
Ready to get started with Databricks Mosaic AI Agent Framework?
View Pricing Options →Databricks Mosaic AI Agent Framework works with these platforms and services:
We believe in transparent reviews. Here's what Databricks Mosaic AI Agent Framework doesn't handle well:
Through 2025 and into 2026 Databricks has continued to expand Agent Bricks with automated optimization for instructed retrievers and document intelligence agents, reducing the manual tuning previously required for production RAG. Mosaic AI Agent Evaluation has matured with broader LLM-as-judge metric coverage and tighter integration with human review apps for SME labeling. The platform has deepened support for governed tool calling through Unity Catalog functions, enabling agents to invoke SQL, Python, and external APIs under the same permission model as data access. AI Gateway has added more granular guardrails (PII detection, content safety) and broader third-party model coverage. DBRX and Llama family models remain first-class options on Foundation Model APIs alongside expanded partner model access via the gateway.
Pricing and features change often. If anything on this Databricks Mosaic AI Agent Framework page is wrong or missing, tell us and include a link to the source. We review every correction.