Real-time conversational analytics with zero data movement. Bridge your AI to live sources without the ETL tax.
Ask questions in plain English. PlyDB lets your AI agent query live data and return real answers.
PlyDB works with any AI agent that supports MCP or CLI tools — including Claude, ChatGPT, Gemini, Codex, and more.
When your AI agent has a secure, real-time view of your data, it can answer complex questions the moment you ask them.
PlyDB connects your AI simultaneously to app logs in S3, your production database, and your codebase — so it can trace an incident from symptom to root cause and draft fixes in a single session, without manual data extraction.
"Review our app logs in S3 for errors from the past week. Cross-reference with our codebase and database replica to identify affected customers and diagnose root causes. Open PRs with fixes for the most severe issues, and draft a summary our PM and CSM teams can use."
PlyDB joins CRM records, support tickets, and product usage data in one query — giving your AI a complete account picture to identify churn risk and generate personalized outreach, without exporting anything to a spreadsheet.
"Analyze our top 20 accounts by revenue. Cross-reference their support tickets with recent product usage. Generate a churn-risk dashboard and draft personalized outreach for the three accounts with the lowest activity."
PlyDB joins signed contracts in a Google Sheet against live seat counts in your database — so your AI can surface billing gaps and calculate unbilled revenue in real time, without a manual audit.
"Audit our active seat counts against signed contracts in the Google Sheet. Identify all overages, calculate the total unbilled revenue, and build a summary table the billing team can use to issue invoices."
PlyDB federates CRM data, infrastructure cost logs, and HR spreadsheets into a single query — so your AI can build a cross-functional financial picture that would otherwise require three separate exports and a BI tool.
"Pull the MRR from the CRM, infrastructure costs from our logs, and headcount from the HR spreadsheet. Build a financial health dashboard and suggest three areas where we can improve our operating margin."
PlyDB is an open-source universal database gateway for AI agents. It sits between your agent and your data sources — databases, flat files, object storage, and SaaS — executing queries directly against each source in real time. No data movement. No pipeline to build. No warehouse to maintain.
Raw schemas tell an agent what columns exist. Semantic context tells it what they mean. PlyDB bridges that gap automatically — and lets agents deepen their understanding over time.
PlyDB scans your data sources and generates structured semantic context — schemas, tables, columns, types, and database COMMENT metadata — all without any manual setup.
Output follows the open OSI standard, giving agents a machine-readable map of your data model.
Schemas describe structure, but agents can learn meaning from richer sources — your codebase reveals enum values, validation rules, and business logic, while your conversations teach domain context that no schema can express.
An agent that sees status = 3 can check your code to learn it means “churned” — or remember that you explained it last Tuesday.
Every insight — whether discovered from schemas, learned from source code, or explained by a human — can be recorded into semantic overlay files that persist across sessions and agents.
Overlays follow the open OSI standard and compound over time. Future agents inherit institutional knowledge from day one.
The result: agents that ask better questions, write more accurate queries, and deliver answers you can trust.
Single binary, no dependencies. Install on your machine and connect to live data with zero infrastructure overhead.
Your AI can analyze and find patterns, but it cannot alter your original data unless you explicitly allow it.
Integrate with any AI agent via the command line or Model Context Protocol. Works with Claude, ChatGPT, Gemini, and more.
No pipelines. No staging tables. No warehouse. Query your data exactly where it lives — in real time.
JOIN a PostgreSQL table with a CSV file and a Google Sheet in a single SQL query. PlyDB handles the federation.
Apache 2.0 licensed. No vendor lock-in, full transparency, and a growing community of contributors.
One relational interface across databases, flat files, object storage, and SaaS platforms.
PlyDB is an open-source universal database gateway for AI agents. It connects your AI to live data sources — PostgreSQL, MySQL, SQLite, DuckDB, CSV, JSON, Parquet, Excel, Amazon S3, and Google Sheets — so it can query across all of them using standard SQL, with no ETL pipeline or data warehouse required.
PlyDB sits between your AI agent and your data sources. You ask a question in plain English; your agent translates it into SQL and sends it to PlyDB via MCP or CLI. PlyDB executes the query directly against your connected sources and returns the results — no data movement, no pipeline.
PlyDB works with any AI agent that supports MCP (Model Context Protocol) or CLI tools — including Claude, Claude Code, Claude Desktop, ChatGPT, Gemini, Codex, and more.
PlyDB currently supports PostgreSQL, MySQL, SQLite, DuckDB, CSV, JSON, Parquet, Excel, Amazon S3, and Google Sheets. Apache Iceberg and Delta Lake support are planned.
No. Your AI agent handles all the SQL. You ask questions in plain English; the agent translates them into queries, runs them via PlyDB, and returns answers.
No. PlyDB runs locally on your machine and queries your sources directly. Your data never leaves your infrastructure — only query results are returned to your AI agent.
A data warehouse requires copying and transforming your data into a central store before you can query it — adding cost, latency, and maintenance. PlyDB queries your data exactly where it lives, in real time, with no pipeline and no data movement.
Yes. PlyDB is open source under the Apache 2.0 license — no vendor lock-in, full transparency, and a community-driven codebase.
More questions? See the full FAQ in the docs.
Install PlyDB with a single command. No runtime dependencies, no containers, no additional infrastructure required.
$ curl -fsSL https://raw.githubusercontent.com/kineticloom/plydb/main/install.sh | shPS> irm https://raw.githubusercontent.com/kineticloom/plydb/main/install.ps1 | iexQuery every source with standard SQL. No pipelines. No warehouse. No lock-in.
Apache License 2.0