Portia

Portia is an AI copilot for data work that profiles tables, answers questions, and builds dbt SQL.

Portia is an AI copilot for data work that connects to Snowflake, BigQuery, PostgreSQL, or local CSV and Parquet files. It profiles tables, answers questions, draws charts, and builds tables one decision at a time, recording each in a YAML spec that compiles to dbt-shaped SQL. The model never reads your data directly; every number is computed by deterministic code to ensure accuracy. It uses Neo4j for a knowledge graph that tracks column lineage and measured overlaps. The copilot runs on the Claude Agent SDK and drives Claude Code, or runs on local models served by Ollama or llama.cpp.

Portia at a glance

Best for: Software engineer, Researcher
Runs on: CLI, self-hosted
Pricing: Free
Released: October 2026
Category: AI research tool
Website: github.com/jad1908/portia

Key features

Connects to Snowflake, BigQuery, PostgreSQL and local files
Profiles data and diagnoses joins before execution
Renders interactive charts in the conversation tab
Compiles YAML decision specs into dbt-shaped SQL
Maintains a knowledge graph for column lineage in Neo4j
Records all chats, indexing jobs, and data questions in a journal
Runs locally using Ollama or llama.cpp models

Use it for

Profile database tables and local file folders
Ask questions about data and generate charts
Build data pipelines compiled to dbt SQL
Check data drift against recorded YAML decision specs

Alternatives to Portia

Portkey — Unified API access with observability, guardrails, and caching for LLM applications.
DBX — It combines lightweight size, open-source access, and built-in AI for over one hundred databases.
DB-GPT — Builds autonomous agents for complex data tasks with private LLM technology.
ktx — It gives data agents a shared context workspace of approved metrics and git-based files.
Lacuna by Tiptree Systems — Traces code imports and comments directly to the foundational research papers.
Decido — Runs locally on CPU while beating published benchmark scores.
Vanna AI — It scales from prototype to multi-tenant production without rewrites.
AskYourDatabase — Query databases, inspect generated SQL, and build charts using natural language.
FAROS — Connects literature evidence, executable plans, and human review into an auditable research loop.
Vizro — Combines low-code configuration with high-code customization for data dashboards.
yggy — Keeps knowledge graphs automatically up to date with source citations for every fact.
Cube — It keeps metric definitions consistent across AI agents and human users.

All Portia alternatives