# TiDB Messaging Playbook

> Approved messaging snapshot — the output of the latest `tidb-messaging-review` PASS.  
> All external content must align with the messaging in this file.  
> **Replace with a new snapshot after each messaging review cycle.**

Last messaging review: 2026 (AI-Native era — updated with competitive intelligence from 5 comparison docs)  
Status: Draft (pending formal `tidb-messaging-review` PASS — update status when review completes)

---

## Tier 1: Core positioning

**One-sentence positioning:**
> TiDB is the AI-native distributed SQL database that unifies vectors, transactions, and analytics in one engine, enabling AI application builders to scale agent workloads without managing separate data systems.

**Lead tagline:**
> Scale Your AI Agents Without Compromise

**Category frame:**
AI-Native Distributed SQL Database

**Key differentiation (2026 competitive context):**
The market has fragmented into specialized systems — vector DBs, OLTP databases, analytics warehouses, and AI tooling — each solving one problem. TiDB is the consolidation play: one SQL engine handles all of it. This matters most when AI workloads need ACID consistency across vectors and transactions, or when teams are tired of maintaining ETL pipelines between RDS + ClickHouse + Elasticsearch stacks.

---

## Tier 2: Segment value propositions

### AI Agent Builders
> "Build AI agents that maintain consistent state across vector retrieval and transactional writes — in a single database, without a sync pipeline."

**Proof point:** Manus migrated 1M+ database tenants in 2 weeks. Dify consolidated backend infrastructure unifying vectors, documents, and relational data.  
**Key capability:** Native HNSW-based vector indexing + ACID transactions + SQL hybrid queries  
**Database branching for agent sandboxing:** Agents can create isolated branches to validate generated SQL and simulate "what-if" scenarios without touching production data.  
**Entry motion:** TiDB Cloud Starter (free, scales to zero) → SLG upgrade at scale

### AI SaaS / Multi-Tenant Platforms
> "Scale from 10 to 10,000 customers without scaling your ops team — tenant isolation without per-tenant clusters."

**Proof point:** Atlassian consolidated 750+ Postgres clusters to 16 TiDB clusters.  
**Key capability:** Workload isolation, resource quotas, database branching  
**Competitive context:** When teams outgrow Supabase (single-node Postgres) or Neon (vertical-only, now Databricks-controlled), TiDB Cloud Starter provides the same free-tier entry with a path to distributed scale without a full migration.

### Scale-Out MySQL
> "MySQL protocol compatible, horizontally scalable — migrate from MySQL or Vitess without rewriting your application."

**Proof point:** Plaid migrated 96 Aurora services. 52–67% cost savings vs RDS MySQL at 8,000 QPS / 500GB.  
**Key capability:** Transparent sharding, distributed ACID, MySQL compatibility  
**Competitive context:** Aurora hits the single-writer ceiling; Vitess adds operational complexity. RDS for MySQL has blocking schema changes and no zero-downtime major upgrades. TiDB has non-blocking DDL and rolling upgrades from day one.

### HTAP / ETL Elimination
> "Query last 30 seconds of transactions analytically — TiFlash reads from the same data as TiKV. No ETL, no pipeline, no staleness."

**Key capability:** TiFlash columnar engine, cost-based query routing  
**Competitive context:** Teams running RDS + ClickHouse + Elasticsearch store the same data 3+ times and maintain Kafka/Debezium pipelines. If that ETL pipe breaks, analytics and search go stale. TiDB eliminates the pipeline — one query can access Row and Column stores simultaneously.  
**Honesty disclosure:** TiFlash replication lag is typically sub-second to a few seconds (not instantaneous); disclose this when relevant.

### Fintech / High-Stakes OLTP
> "Financial-grade ACID transactions at distributed scale — multi-region active-active, 99.999% availability SLA."

**Proof point:** Plaid (99.99% uptime, 96 Aurora services migrated), Chainbase (50% infrastructure cost reduction), SB Payment Service (zero-downtime scaling).  
**Key capability:** Distributed transactions, multi-region, HA  
**Competitive context:** RDS multi-region failover has up to 35-second delay. TiDB uses Raft-based consensus for near-zero RPO.

### Developer Entry (TiDB Cloud Starter)
> "Start free. Scale automatically. No cluster management — not even when you hit enterprise scale."

**Key capability:** Scales to zero, pay-as-you-go, 25 GiB free storage, non-blocking schema changes, vector search, AI-assisted SQL editor  
**Competitive context:** PlanetScale requires $3,000/mo enterprise plan for auto-scaling and multi-region. TiDB Cloud Starter includes both on the free tier.

---

## Tier 3: Approved taglines by use case

| Use case | Tagline |
|---------|---------|
| Brand / general | Scale Your AI Agents Without Compromise |
| AI workload | One database. Vectors, transactions, analytics. |
| Scale-out MySQL | MySQL protocol compatible. Horizontally scalable. |
| HTAP | Transact and analyze. Same data. No ETL. |
| Developer entry | SQL you know. Scale you need. |
| Enterprise | Mission-critical SQL at any scale. |
| vs. siloed stack | One engine. No pipelines. |

**Advisory on "MySQL compatible. Infinitely scalable.":** Remove "infinitely" — use "horizontally scalable" per banned phrases list.

---

## Tier 4: Proof points (approved for external use)

| Customer | Claim | Permission | Notes |
|---------|-------|-----------|-------|
| Manus | 1M+ database tenants, 2-week migration | Public | AI agent platform use case |
| Dify | Consolidated backend; unified vectors, documents, relational data | Public | AI workflow platform |
| Atlassian | 750+ Postgres clusters → 16 TiDB clusters | Public | Enterprise consolidation |
| Plaid | 96 Aurora services migrated; 99.99% uptime | Public | Fintech OLTP |
| Chainbase | 50% infrastructure cost reduction | Public | Web3 / fintech |
| Pinterest | 10x P99 latency improvement | Public | Graph service workload |
| ZTO Express | 25% reduction in per-order costs | Public | Logistics / e-commerce |
| Catalyst | 60x faster performance | Public | Operational intelligence |
| Square, Bolt | Logo display | Display only | No specific metrics approved |

---

## Tier 5: Banned phrases (Tier 1 — block)

- "fully MySQL compatible" → use "MySQL protocol compatible" with docs link
- "zero downtime" as a generic claim → for TiDB Cloud managed tiers, "zero-downtime upgrades" is valid in comparisons vs RDS; for TiDB Self-Managed, use "rolling upgrade with minimal per-node disruption"
- "seamless", "frictionless", "effortless"
- "game-changing", "revolutionary", "industry-leading", "best-in-class"
- "AI-ready" without specific capability named
- "scales infinitely" → use "scales horizontally to [tested configuration]"
- "real-time analytics" without disclosing TiFlash replication lag (typically sub-second to a few seconds)
- Any benchmark claim without methodology
- Customer name with specific metric without written approval
- "TiDB Serverless" / "TiDB Cloud Serverless" (deprecated) → use "TiDB Cloud Starter"
- "vector database" as a standalone category for TiDB (TiDB is a unified system)

---

## AI integration messaging

**MCP server:** "TiDB's MCP server lets AI agents query and update TiDB using natural language instructions — standard SQL under the hood, agent-native interface."

**pytidb:** "Python SDK for AI workloads — vector search, hybrid queries, and connection pooling optimized for high-concurrency agent applications."

**LangChain / LlamaIndex:** "TiDB integrates as a vector store in LangChain and LlamaIndex — load once, query with SQL alongside your vectors."

**Database branching for AI:** "Agents can create isolated database branches within seconds for sandboxed experiments — validate generated SQL, run what-if scenarios, spin up tenant environments without touching production."

---

## Competitive messaging angles (2026)

| Competitor context | TiDB message |
|-------------------|--------------|
| vs. Neon (now Databricks) | "Neon routes analytics to Databricks, vectors to Pinecone, and OLTP to Neon — three systems, two ETL hops. TiDB handles all three in one engine." |
| vs. Supabase at scale | "Supabase is excellent for MVPs. When you outgrow single-node Postgres, TiDB Cloud Starter has the same free entry with a path to distributed scale — no full migration required." |
| vs. PlanetScale | "PlanetScale's auto-scaling costs $3,000/mo. TiDB Cloud Starter auto-scales from zero — free." |
| vs. RDS + ClickHouse + ES | "You didn't choose three databases — your stack grew by accident. TiDB replaces the spaghetti architecture with one SQL engine and zero ETL." |
| vs. Aurora | "Aurora hits the single-writer ceiling. TiDB's distributed writes scale horizontally. And schema changes don't take your database offline." |

---

## Messaging DO / DON'T quick reference

| DO say | DON'T say |
|--------|----------|
| "MySQL protocol compatible" | "MySQL compatible" / "fully MySQL compatible" |
| "Horizontal scale-out" | "Infinite scale" |
| "Native HNSW-based vector indexing" | "Vector database" (TiDB is not primarily a vector DB) |
| "ACID consistency for agent state" | "Best database for AI" |
| "TiFlash for real-time HTAP queries (sub-second to few-second lag)" | "Real-time analytics" without lag disclosure |
| "TiDB Cloud Starter" | "TiDB Serverless" / "TiDB Cloud Serverless" (deprecated) |
| "Database branching for agent sandboxing" | "Seamless agent isolation" |
| "Snapshot Isolation / REPEATABLE READ by default" | Implying SERIALIZABLE isolation by default |