Agentic Postgres

Agentic Postgres is a PostgreSQL database purpose-built for AI agents and Developers. It combines MCP integration, vector embeddings, hybrid search, time-series analytics, and observability in a single, intelligent platform.

Get Started in 3 Commands

Install and start building with Agentic Postgres

# Install Tiger CLI
$ curl -fsSL https://cli.tigerdata.com | sh
# Authenticate
$ tiger auth login
# Install MCP
$ tiger mcp install
Start for freeGitHub Demo

Free tier available. No credit card required.

Why Agentic Postgres?

Users want experiences like ChatGPT. They expect apps to intelligently figure out what tools are needed, think through problems step by step, and adapt to their questions and tasks in real time. They no longer want traditional apps with rigid workflows. They want agentic apps where agents translate plain English queries (whether it's a question or a task) into action. This means the entire tech stack needs to be agentic, including the database.

At the same time, developers also want experience like ChatGPT. They want to “vibe code” using agentic tools and technologies like MCP, Cursor, Claude, and GPT to build their applications. It's not just about writing code. They also want to talk to their database in plain English for schema design, analyzing slow queries, and expect the agents to handle those tasks.

To make this work, your database needs to adapt as well. A simple question in plain English, whether from an end user or a developer, can trigger complex workflows behind the scenes. The database needs to support these workflows intelligently.

During Development:

When a developer asks “Should I add an index here?” or “Why is this query slow?”, the database, through MCP (Model Context Protocol), becomes an active partner. It uses 35+ years of combined experience and searches official Postgres documentation, analyzes query plans, suggests optimal table structures, and recommends indexes. It's like having a database expert embedded in your IDE. Need to test a new schema or index? Create an instant fork of your entire database in seconds with zero-copy technology. Test your changes in a safe, isolated sandbox, then merge back or discard - all without touching production data.

During Production:

When an end user asks “Why did sales drop last month?” or “Are customers complaining about our new feature?”, that simple question triggers dozens of sophisticated database operations:

  • →Multiple regular SQL queries for structured data analysis
  • →Time-series queries to identify trends and anomalies over time
  • →Keyword search queries to find exact mentions and phrases
  • →Semantic search queries to understand meaning and sentiment
  • →Hybrid search queries combining multiple techniques for maximum recall
  • →Memory updates to store insights and maintain context across investigations

To support this, Postgres needs to have all of these capabilities built in. MCP for intelligent interactions. Vector embeddings for semantic search. BM25 for keyword search. Hybrid search combining both. Time-series analytics. All working together, at scale, faster than stitching together multiple specialized databases.

That's why TigerData, bundled all these capabilities into one database. That's Agentic Postgres.

The Building Blocks

Essential capabilities working together as one intelligent database

🤖

Smart MCP Server

Built-in MCP servers connect agents to Postgres and TigerData docs. Master prompts teach agents how to design schemas, tune queries, and reason with context so they build smarter, not just faster.

Read the docs
🔍

Native AI Search

Postgres now speaks AI search. BM25 text search and vector retrieval run natively inside the database, giving agents fast, high-recall context without external stores or latency overhead.

Read the docs
⚡

Fast, Zero-Copy Forks

Create full Postgres clones in seconds with copy-on-write storage. Agents get fast, isolated sandboxes to test, debug, and iterate safely without duplicating data or touching production.

Read the docs
🧠

Semantic Search

Understand meaning, not just keywords. pgvectorscale delivers high-throughput similarity search with better recall than pgvector at scale.

Read the docs
🎯

Hybrid Search

Combine BM25 and vector search in a single query. Get the best of both worlds: keyword precision and semantic understanding.

Read the docs
⏱️

Time-Series Native

TimescaleDB hypertables for time-series data. Automatic partitioning, compression, and continuous aggregates built in.

Read the docs

= One Database, All Capabilities, Infinite Intelligence

See Agentic Intelligence in Action

Watch how an agent investigates a business problem using plain English queries.
Each question builds on the last, showcasing different capabilities.

1

“Why did sales drop?”

CAPABILITY:
TimescaleDB + Multi-Query Analysis

What it does:

AGENTIC multi-step reasoning

AGENT RESPONSE

Analyzing sales data across time series... Found significant drop in Product X sales starting March 15th. Correlating with events... Detected quality issues reported around the same timeframe.

SQL QUERY EXAMPLE
SELECT product_id, time_bucket('1 day', sale_date) as day, sum(revenue) FROM sales GROUP BY product_id, day ORDER BY day DESC;
2

“Are customers complaining?”

CAPABILITY:
pgvectorscale + Semantic Search

What it does:

Semantic search finding hidden complaints

AGENT RESPONSE

Searching customer feedback using semantic similarity... Found 47 complaints about Product X, including phrases like 'not working as expected', 'quality concerns', and 'disappointed with recent purchase'.

SQL QUERY EXAMPLE
SELECT * FROM feedback WHERE embedding <-> embedding_of('product quality issues') < 0.3 ORDER BY created_at DESC;
3

“Find exact mentions of 'broken'”

CAPABILITY:
pg_textsearch (BM25)

What it does:

Full-text search

AGENT RESPONSE

Executing BM25 full-text search... Found 23 exact matches for 'broken' in customer feedback. Most mentions reference Product X components failing within 2 weeks of purchase.

SQL QUERY EXAMPLE
SELECT * FROM feedback WHERE to_tsvector('english', content) @@ to_tsquery('broken') ORDER BY ts_rank(to_tsvector('english', content), to_tsquery('broken')) DESC;
4

“Find ALL quality feedback”

CAPABILITY:
BM25 + Vector Combined

What it does:

HYBRID search (the hero!)

AGENT RESPONSE

Combining full-text and semantic search for maximum recall... Found 89 quality-related feedback items (67 from semantic search, 23 from keyword search, 1 from both). Product X accounts for 78% of all quality complaints.

SQL QUERY EXAMPLE
SELECT *, (bm25_score * 0.5 + similarity_score * 0.5) as hybrid_score FROM feedback WHERE to_tsvector('english', content) @@ to_tsquery('quality') OR embedding <-> target_embedding < 0.3 ORDER BY hybrid_score DESC;
5

“Connect the dots”

CAPABILITY:
Cross-Query Synthesis + pgai

What it does:

Memory retrieval across investigations

AGENT RESPONSE

Synthesizing insights from all investigations... Timeline: Sales dropped March 15th → Quality complaints increased → 'Broken' mentions spiked → 89 total quality issues identified. Root cause: Product X manufacturing defect affecting 78% of recent batches. Recommendation: Issue recall and investigate supplier.

SQL QUERY EXAMPLE
SELECT * FROM investigation_memory WHERE session_id = current_session ORDER BY timestamp; -- AI synthesizes patterns
6

“How did you do this?”

CAPABILITY:
It's All One Database!

What it does:

Self-observability - THE BIG REVEAL

AGENT RESPONSE

🎯 Everything you just saw happened in ONE Postgres database: ✅ Time-series queries (TimescaleDB) ✅ Semantic search (pgvectorscale + pgai) ✅ Full-text search (pg_textsearch BM25) ✅ Hybrid search (combining both) ✅ Memory & synthesis (stored with embeddings) ✅ Self-analysis (querying my own history) No external services. No data movement. Just intelligent Postgres.

SQL QUERY EXAMPLE
SELECT DISTINCT extension_name FROM pg_extension WHERE extension_name IN ('timescaledb', 'vector', 'vectorscale', 'ai', 'pg_textsearch', 'plpgsql');
🎯

The Magic: It's Just Postgres

No external APIs. No data pipelines. No complexity.
One database. Six extensions. Infinite intelligence.

The Tech Stack

Six powerful extensions working together as one. Replace your entire data infrastructure with intelligent Postgres.

timescaledb

time-series

Hypertables, time-series functions

Replaces: InfluxDB, Prometheus

vector

ai

Vector data type, basic ops

Replaces: Foundation layer

vectorscale

ai

DiskANN index for fast similarity

Replaces: Pinecone, Weaviate

ai (pgai)

ai

Generate embeddings in DB

Replaces: Separate embedding service

pg_textsearch

Built-in

BM25 keyword search

Replaces: Elasticsearch, Algolia

plpgsql

Built-in

Procedural language for logic

Replaces: Application-layer logic

Replace Your Entire Stack

From five separate databases to one intelligent system

Before: Traditional Stack
PostgreSQLInfluxDBElasticsearchPineconeOpenAI API

Five services • Data synchronization • Higher costs • Operational complexity

After: Agentic Postgres
Agentic Postgres

One database • Zero data movement • Simplified operations • Lower total cost

Get Started in 3 Commands

Install the Tiger CLI and MCP. Create your first free service. Start building.

# Install Tiger CLI
$ curl -fsSL https://cli.tigerdata.com | sh
# Authenticate
$ tiger auth login
# Install MCP
$ tiger mcp install
Read the Launch PostTry Free Tier

Built for Agents. Designed to Elevate Developers.

Agents are the new developers. Agentic Postgres is their new playground. Join thousands of developers building the future with AI.