Azure Cosmos DB's Visual Studio Code extension now integrates GitHub Copilot tools directly into the Query Editor, letting Copilot inspect connection context, sample container schemas, generate schema-grounded NoSQL queries, and execute them with explicit developer consent. Dedicated agent skills teach Copilot Cosmos DB-specific query dialect and over 100 best-practice recommendations across data modeling, partitioning, and indexing. An optional MCP server mode via Azure Cosmos DB Shell extends these capabilities to longer-running agent workflows outside the editor, while the Linux-based vNext Emulator supports local and CI testing with synthetic data. A preview AI-assisted Migration Assistant helps teams move relational workloads to Cosmos DB, and an Account Overview dashboard offers read-only operational visibility. Throughout, human developers retain control over data access, RU costs, and execution approval, with existing Cosmos DB identity and permission models governing what agents can do.
Table of contents
A query is more than query text Copy linkCheck the schema before guessing Copy linkTools still need database judgment Copy linkKeep a person in the loop Copy linkWhen the Query Editor isn’t the whole workflow Copy linkStart with fake data and a local database Copy linkStay in the flow Copy linkBefore the first query and after the thousandth Copy linkThe point isn’t autonomy Copy linkAbout Azure Cosmos DB Copy linkQuestions this post answers
How does the Azure Cosmos DB VS Code extension let GitHub Copilot generate accurate NoSQL queries without guessing property names?
Copilot can sample the active container's schema before generating a query, inferring real property names and types instead of guessing based on the prompt alone. This prevents errors like assuming a property is named isActive when the actual field is accountStatus. Schema sampling requires developer consent because it reads data and consumes request units, and the generated query lands in the Query Editor for review before execution. Developers wiring AI agents into database workflows can track tooling changes like this through daily.dev.
What is the Azure Cosmos DB Agent Kit and which AI coding tools does it support?
The Azure Cosmos DB Agent Kit is a skill set covering more than 100 recommendations across data modeling, partition-key design, query optimization, SDK usage, indexing, throughput, global distribution, monitoring, vector search, and full-text search. It works across GitHub Copilot, Claude Code, Codex, Cursor, Gemini CLI, and other compatible agents, giving them Cosmos DB-specific knowledge during everyday coding without repeatedly pasting documentation into conversations. Anyone comparing agent skill kits across coding assistants can follow updates like this on daily.dev.
How can I test AI agent database queries locally before running them against production Cosmos DB data?
The Linux-based vNext Azure Cosmos DB Emulator runs in Docker and includes Data Explorer and Azure Cosmos DB Shell, letting developers create realistic containers with synthetic data that has missing fields and inconsistent shapes to stress-test schema sampling. It supports the API for NoSQL in gateway mode but only a subset of cloud capabilities, so request-unit behavior and some production features aren't fully represented, meaning performance and scale still need validation in Azure. Teams building safe local testing pipelines for agent-driven database work can keep up via daily.dev.