Tidb
TiDB is built for agentic workloads that grow unpredictably, with ACID guarantees and native support for transactions, analytics, and vector search. No data silos. No noisy neighbors. No infrastructure ceiling.
Last verified:
What is Tidb?
TiDB is an open-source, cloud-native, distributed SQL database purpose-built for AI agents and modern applications that require elastic scale, transactional consistency, and unified data access. It combines vectors, transactions, and analytics in a single engine with ACID guarantees, eliminating the need for separate vector stores, data warehouses, and transaction layers. TiDB enables developers to join vector similarity results directly against transactional data in one query without ETL or middleware.
Key features include automatic autoscaling that adapts from zero to peak demand and scales to zero when idle, instant branching for isolated environments in milliseconds, and independent compute/storage scaling without capacity planning. TiDB maintains ACID-consistent agent state across workflow steps, ensures concurrent agents read identical state regardless of node, and supports multi-step workflows that survive failover. It offers native vector search with vec_cosine_distance functions, HTAP capabilities for hybrid transactional/analytical processing, MySQL compatibility, and integrations with MCP, pytidb, LangChain, and LlamaIndex.
TiDB is designed for AI developers building agentic workloads, enterprises needinghorizontal scalability, SaaS companies with variable workloads, and teams running concurrent AI agents requiring tenant-scoped data separation. It serves finance, gaming, and e-commerce sectors where real-time insights are crucial. The database is ideal for applications requiring high availability, multi-region deployment, and zero-downtime schema changes.
Tidb pricing
Pricing model: Freemium
TiDB Cloud Starter starts from $0/month with a generous free tier: 25 GiB row storage, 25 GiB column storage, and 250M Request Units (RUs) free per month for each organization (5 GiB row + 5 GiB column + 50M RUs per cluster, up to 5 clusters). Additional storage costs $0.20 per GiB, additional RUs cost $0.10 per 1M RUs. Pay only for what you use with spend limits to avoid overage. TiDB Cloud Essential Preview costs approximately $20/day for typical small production workloads with provisioned compute autoscaling up to 100K RUs, point-in-time backup, and 99.99% availability. TiDB Cloud Premium Preview starts from $1800/month for mission-critical hyper-scale applications with unlimited autoscaling and enterprise security (CMEK, PrivateLink, VPC Peering). TiDB Cloud Dedicated starts from $0.22/hr or $1376/month with 4-32 vCPU per node, PCI-DSS and SOC 2 Type II compliance, available on AWS, Google Cloud, and Azure. TiDB Self-Managed pricing is upon request for advanced configuration requirements on any infrastructure.
Tidb pros
- Elastic autoscaling from zero to peak without capacity planning
- Scale down to zero when idle to save costs
- Instant branching creates isolated environments in milliseconds
- Unified engine replaces vector store, warehouse, and transaction layer
- Join vector similarity directly against transactional data in one query
- No ETL or middleware required for unified data access
- ACID-consistent state across multi-step agent workflows
- Concurrent agents always read identical state on any node
- Multi-step workflows survive failover without losing progress
- Dedicated throughput per agent without noisy neighbor interference
- Tenant-scoped data separation keeps agent state private
- Standard SQL means developers productive from day one
- Native MCP, pytidb, LangChain, and LlamaIndex integrations
- Online DDLs and zero-downtime schema changes
- Built-in vector search with vec_cosine_distance function
- HTAP capabilities blend transactional and analytical workloads
- 99.99% availability with multi-zone protection
- MySQL protocol compatible for easy migration
- Point-in-time backup with up to 30-day retention
- Compute and storage scale independently
Tidb cons
- Writing performance limited for large single-table writes
- Complex analytical queries slower than dedicated OLAP databases like ClickHouse
- Deployment and tuning requires distributed database experience
- Higher memory consumption than traditional single-node databases
- Ecosystem and tool support less mature than MySQL or PostgreSQL
- Some advanced MySQL features not fully supported yet
- TiFlash setup and resource planning tricky at scale
- Learning curve around distributed architecture and performance tuning
- Free tier limited to 256 MiB memory per query
- Not yet available on Google Cloud or Azure for Starter tier
Frequently asked questions about Tidb
What is TiDB Cloud Starter?
TiDB Cloud Starter offers the TiDB database with full HTAP capabilities for you and your organization. It is a fully managed, auto-scaling deployment of TiDB that lets you start using your database immediately, develop and run your application without caring about the underlying nodes, and automatically scale based on your application's workload changes. It was previously called TiDB Cloud Serverless, renamed effective August 12, 2025 to make the entry tier's purpose clearer.
Is TiDB Cloud Starter free to use?
Yes, TiDB Cloud Starter provides an always-free monthly quota for up to five clusters per organization. Each cluster gets 5 GiB row-based storage, 5 GiB columnar storage, and 50 million Request Units (RUs) per month. For the first five clusters in your organization, you get a total of 25 GiB row storage, 25 GiB column storage, and 250M RUs free per month. No credit card is required to get started with the free tier.
What are Request Units (RUs)?
Request Units are an abstract measurement for the size and intricacy of requests initiated on your TiDB Cloud Starter instance. All activities such as SQL queries, bulk operations, and background jobs are quantified in RUs. TiDB Cloud Starter adopts a pay-as-you-go model where you only pay for storage space and instance usage beyond the free quota. Additional RUs cost $0.10 per 1M RUs. You can use EXPLAIN ANALYZE to get RU consumption of individual SQL statements.
Can I use TiDB Labs in production?
TiDB Cloud Starter is production-ready and serves as the on-ramp for thousands of developers who need a production-ready database that scales automatically. However, TiDB Labs itself is for experimentation and learning. For production workloads needing performance, security, and scale without infrastructure management, you can use TiDB Cloud Starter or upgrade to TiDB Cloud Essential Preview or Premium Preview for mission-critical applications.
How many TiDB Cloud Starter instances can I create?
For each organization in TiDB Cloud, you can create a maximum of five TiDB Cloud Starter instances by default. To create more than five instances, you need to add a credit card and set a monthly spending limit for the usage. If you delete some previous instances before creating a new one, the new instance will still have the free quota applied.
What is columnar storage in TiDB Cloud Starter?
Columnar storage acts as an additional replica of row-based storage with strong consistency. Unlike row-based storage that stores data in rows, columnar storage organizes data in columns, optimizing it for analytics tasks. It enables HTAP capabilities by blending transactional and analytical workloads. TiDB Cloud Starter uses a separate elastic TiFlash engine to manage columnar storage data. The optimizer automatically decides whether to retrieve data from row-based or columnar storage during query execution.
When should I upgrade from Starter to Essential or Premium?
Upgrade to TiDB Cloud Essential Preview for production workloads needing performance, security, and scale without infrastructure complexity, with provisioned compute autoscaling up to 100K RUs and typical small prod costs around $20/day. Upgrade to Premium Preview for mission-critical, hyper-scale applications demanding unlimited growth, enterprise-grade security (CMEK, PrivateLink, VPC Peering), starting from $1800/month. Upgrade to Dedicated for predictable traffic with advanced requirements on AWS, Google Cloud, or Azure, starting from $0.22/hr.
What happens when I exceed my free quota?
Once a TiDB Cloud Starter instance reaches its free quota or spending limit, the instance immediately denies any new connection attempts until the quota is increased or usage resets at the start of a new month. Existing connections established before reaching the quota remain active but experience throttling. For free instances, once the free quota is reached, read and write operations are throttled until you set a monthly spending limit or the usage resets monthly.
Is TiDB Cloud Starter available on Google Cloud or Azure?
TiDB Cloud Starter is currently available on AWS and Alibaba Cloud. The team is actively working on expanding to Google Cloud and Azure, but there is no exact timeline as they focus on filling gaps and ensuring seamless functionality across all environments. TiDB Cloud Dedicated is available on AWS, Google Cloud, and Azure for applications with predictable traffic and advanced requirements.
How does TiDB handle AI agent workloads?
TiDB is purpose-built for agentic scale with transactional speed, unified data access, and elastic scale that AI agents need to reason, act, and iterate in real time. Each agent gets dedicated throughput that other workloads can't touch, with tenant-scoped data separation keeping every agent's state private. TiDB maintains ACID-consistent agent state from one workflow step to the next, ensures concurrent agents always read identical state regardless of node, and supports multi-step workflows that survive failover without losing progress or replaying actions. It integrates with MCP, pytidb, LangChain, and LlamaIndex out of the box.