# Powabase - [Platform Overview](https://docs.powabase.ai/concepts/platform-overview.md): Powabase is a fully managed backend (database, auth, storage, and AI) that your coding agent can build on top of. Point Claude Code, Cursor, Copilot, or an AI app builder like Replit, Base44, or Lovable at your project, and ship a real app without standing up any infrastructure. - [Auth & Connection](https://docs.powabase.ai/guides/auth-connection.md): Find your project's credentials in the Connect modal, see which key each /api route accepts, and make your first authenticated request. - [Quickstart](https://docs.powabase.ai/guides/quickstart.md): Build an end-to-end RAG agent in 5 minutes, from document upload to a streaming conversation. - [Platform Comparison](https://docs.powabase.ai/concepts/platform-comparison.md): How Powabase compares to popular AI frameworks, RAG services, workflow tools, and backend platforms. Where each tool excels, and how Powabase's unified approach differs. - [Sources & Extraction](https://docs.powabase.ai/concepts/sources-extraction.md): Sources are the entry point for documents in the platform. Upload files, and the extraction pipeline converts them into structured text that can be indexed into knowledge bases. - [Knowledge Bases & Indexing](https://docs.powabase.ai/concepts/knowledge-bases-indexing.md): The platform's Context Engineering suite turns raw documents into searchable knowledge. Choose from five indexing strategies and four retrieval algorithms to build the right pipeline for your use case, whether that's simple vector search or LLM-driven document reasoning. - [Agents & Tools](https://docs.powabase.ai/concepts/agents-tools.md): Agents are LLM-powered conversational entities that use a ReAct loop to reason, call tools, and respond. They support eight builtin tools, custom HTTP tools, MCP server integration, session-based memory, hook-based middleware, and a human-in-the-loop approval flow. - [Multi-Agent Orchestration](https://docs.powabase.ai/concepts/orchestrations-concept.md): Orchestrations coordinate multiple agents to solve complex tasks. Three execution strategies (Supervisor, Sequential, and Parallel) give you different patterns for multi-agent collaboration, from autonomous delegation to pipeline processing to concurrent fan-out with merged results. - [Workflows](https://docs.powabase.ai/concepts/workflows-concept.md): Workflows are DAG-based automation pipelines that chain together LLM calls, code execution, conditions, and agent runs into deterministic, repeatable processes. - [Streaming & SSE](https://docs.powabase.ai/concepts/streaming-patterns.md): The platform uses Server-Sent Events (SSE) for real-time streaming of agent responses, orchestration flows, and workflow executions. This page covers the event format, event types, and consumption patterns. - [Upload Your First Document](https://docs.powabase.ai/guides/upload-document.md): Upload a file (PDF, DOCX, images, etc.) and extract its content. Extracted text becomes available for knowledge base indexing. - [Create a Knowledge Base](https://docs.powabase.ai/guides/create-knowledge-base.md): Index your documents for semantic search and retrieval-augmented generation. Knowledge bases chunk, embed, and store your content for fast vector similarity search. - [Build an Agent](https://docs.powabase.ai/guides/build-agent.md): Create an AI agent, assign tools and knowledge bases, then chat with it via streaming. Agents run a ReAct loop: they reason, call tools, and respond. - [Bring your own LLM](https://docs.powabase.ai/guides/byollm.md): Point agents, copilots, and indexing jobs at any LiteLLM-supported model (including OpenRouter models like DeepSeek) using the correct provider key and model-string format. - [Multi-Agent Orchestration](https://docs.powabase.ai/guides/orchestration.md): Combine multiple agents into an orchestration. A coordinator routes messages to the right agent based on the conversation context. - [Build Workflows Programmatically](https://docs.powabase.ai/guides/workflows-programmatic.md): Create automated workflows by defining blocks and edges via the API. Blocks are processing steps (LLM calls, agent runs, conditions). Edges connect them into a directed graph. - [Build Workflows with Copilot](https://docs.powabase.ai/guides/workflows-copilot.md): Describe what you want in natural language and let the AI copilot build the workflow graph for you. The copilot generates blocks and edges based on your description. - [Streaming Responses](https://docs.powabase.ai/guides/streaming-guide.md): Consume Server-Sent Events from agent runs. Learn to parse events, handle tool calls, manage errors, and build multi-turn conversations. - [Advanced Agent Configuration](https://docs.powabase.ai/guides/advanced-agent-config.md): Configure MCP servers, hooks, and the human-in-the-loop approval flow for production-grade agents. - [Using the Platform with AI Coding Assistants](https://docs.powabase.ai/concepts/ai-coding-assistants.md): Powabase's REST API is built to be consumed by AI coding assistants like Claude Code, GitHub Copilot, Cursor, and others. How to integrate effectively, plus a preview of the upcoming skill framework. - [Sources](https://docs.powabase.ai/api-reference/sources.md): Upload, manage, and extract content from documents and files. - [Knowledge Bases](https://docs.powabase.ai/api-reference/knowledge-bases.md): Create and manage knowledge bases for semantic search and RAG. - [Agents](https://docs.powabase.ai/api-reference/agents.md): Create AI agents, assign tools and knowledge bases, configure MCP servers and hooks, and execute conversations. - [Sessions](https://docs.powabase.ai/api-reference/sessions.md): Manage multi-turn chat sessions and their message/run history. - [Orchestrations](https://docs.powabase.ai/api-reference/orchestrations.md): Combine multiple agents into coordinated multi-agent systems. - [Workflows](https://docs.powabase.ai/api-reference/workflows.md): Create and manage automated block-based workflows with visual graph definitions. - [Webhooks](https://docs.powabase.ai/api-reference/webhooks.md): Trigger deployed workflows from external systems via signed HTTP calls. - [Copilot](https://docs.powabase.ai/api-reference/copilot.md): AI-powered workflow builder. Describe what you want in natural language and the copilot generates the workflow graph. - [Tools](https://docs.powabase.ai/api-reference/tools.md): Manage custom tools and view builtin tools available to agents. - [Context Handlers](https://docs.powabase.ai/api-reference/context-handlers.md): Execute standalone knowledge retrieval outside of agent runs. Useful for building custom RAG pipelines. - [AI Provider Keys](https://docs.powabase.ai/api-reference/ai-provider-keys.md): Store, validate, and rotate per-project credentials for OpenAI, Anthropic, Google, and OpenRouter. - [Settings](https://docs.powabase.ai/api-reference/settings.md): Read and override per-project configuration values managed through a typed registry. - [Architecture](https://docs.powabase.ai/concepts/architecture.md): Understand the control plane / data plane split, per-project isolation, authentication model, and database schemas. - [Database Access](https://docs.powabase.ai/concepts/database-access.md): Every project has a full Postgres database accessible three ways: the Project Service API for AI-managed data, PostgREST for application tables, and a direct Postgres connection for migrations and admin tooling. All three credentials come from the Studio's Connect modal. - [Schemas](https://docs.powabase.ai/concepts/schemas.md): The five schemas you'll find in a Powabase project's Postgres: public (yours), ai (platform-managed), auth (GoTrue), storage (Storage API), and extensions (where pg_net and friends live). - [The ai schema](https://docs.powabase.ai/concepts/ai-schema-postgrest.md): Where the AI surface's state (sources, knowledge bases, agents, runs, sessions, workflows) lives, and why it's backend-only: the ai schema isn't exposed through PostgREST. Reach AI resources through the typed /api/* endpoints. - [AI resource recipes](https://docs.powabase.ai/guides/ai-schema-recipes.md): Worked examples for the jobs people used to do with PostgREST on ai.*: searching several knowledge bases, agent usage summaries, bulk-tagging sources, and per-KB source reports, all through the typed /api/* endpoints. - [Connection Pooling](https://docs.powabase.ai/guides/connection-pooling.md): All external Postgres connections in Powabase go through PgBouncer in transaction mode. What that buys you, what it breaks, and how to write code that works against it. - [Direct Postgres patterns](https://docs.powabase.ai/guides/direct-postgres.md): Connecting to your project's Postgres directly via the pooler URL: psql sessions, prepared statements, transactions, error handling, and the patterns that work safely inside PgBouncer transaction mode. - [Studio SQL Editor](https://docs.powabase.ai/guides/sql-editor.md): Run ad-hoc SQL in the Studio against your project's Postgres: what it's good for, what it's not, the role it runs as, and how to save snippets. - [Migrations](https://docs.powabase.ai/guides/migrations.md): Three migration patterns for evolving your project's schema: hand-written SQL with psql, Prisma Migrate, and Drizzle Kit. Plus the boundary between your tables and the platform's. - [User-managed pgvector](https://docs.powabase.ai/guides/user-pgvector.md): pgvector is preloaded in every Powabase project. When to use the typed AI surface vs roll your own embeddings table, and how to set up HNSW indexes for fast vector search. - [Database webhooks](https://docs.powabase.ai/guides/db-webhooks.md): Fire HTTP requests when rows change, using the supabase_functions.http_request() trigger function plus pg_net. Distinct from the agentic /api/webhooks workflow triggers. - [Prisma](https://docs.powabase.ai/guides/orm-prisma.md): Use Prisma with Powabase's Postgres: connection setup with the pooler-required flags, a simple schema, queries, and migration workflow. - [Drizzle](https://docs.powabase.ai/guides/orm-drizzle.md): Use Drizzle with Powabase: connection setup using postgres.js, schema as TypeScript, queries that look like SQL, and Drizzle Kit migrations. - [SQLAlchemy + Alembic](https://docs.powabase.ai/guides/orm-sqlalchemy.md): Use SQLAlchemy with Powabase: connection setup with psycopg, declarative models, queries, sessions, and Alembic migrations. - [TypeORM](https://docs.powabase.ai/guides/orm-typeorm.md): Use TypeORM with Powabase: DataSource setup for the pooler, decorator-based entities, queries via the repository pattern, and TypeORM migrations. - [Database (Authenticated proxy)](https://docs.powabase.ai/api-reference/database.md): Service-role CRUD over your project's public schema, plus a table introspection endpoint. - [Extensions](https://docs.powabase.ai/api-reference/extensions.md): The Postgres extensions preloaded in every Powabase project, the ones you can CREATE EXTENSION yourself, and where each one lives. - [Database (PostgREST)](https://docs.powabase.ai/api-reference/postgrest.md): Direct REST access to your project's public schema. PostgREST exposes every table and view as a REST endpoint with filtering, ordering, pagination, and embedded relations. - [PostgREST advanced](https://docs.powabase.ai/guides/postgrest-advanced.md): Beyond basic CRUD: embedded joins, full-text search, JSONB operators, counting, upserts, bulk inserts, and the headers that change response shape. - [Row Level Security](https://docs.powabase.ai/concepts/rls-model.md): How Powabase translates an HTTP request's API key or JWT into a Postgres role, and how RLS policies decide which rows that role can see. - [RLS Cookbook](https://docs.powabase.ai/guides/rls-policies.md): Five copy-paste policy patterns for common scenarios: own-rows-only, public-read with auth-write, tenant isolation, role-based access, soft-delete-aware. - [RLS Testing](https://docs.powabase.ai/guides/rls-testing.md): Verify RLS policies in psql or the Studio SQL Editor without spinning up a frontend, by impersonating the anon and authenticated roles with hand-crafted JWT claims. - [Auth model](https://docs.powabase.ai/concepts/auth-model.md): How GoTrue issues JWT access and refresh tokens, what's in each one, the role-based session your app sees afterwards, and how the pieces compose with PostgREST RLS. - [Signup, signin, magic link](https://docs.powabase.ai/guides/auth-signup-signin.md): Working end-to-end auth flows: email + password signup and signin, magic link, password recovery, and session refresh. In Python, TypeScript, and cURL. - [OAuth providers](https://docs.powabase.ai/guides/auth-oauth-providers.md): Wire up Google, GitHub, Apple, and 19 other OAuth providers with PKCE. Configuration, redirect URLs, the sign-in flow, and the gotchas that bite people. - [Auth](https://docs.powabase.ai/api-reference/auth.md): End-user-facing GoTrue endpoints at /auth/v1/*: sign up, sign in, refresh, magic link, password recovery, OAuth, MFA, user management, and admin operations. - [Storage model](https://docs.powabase.ai/concepts/storage-model.md): How Powabase Storage organizes files into buckets, who can access what, what the size and MIME limits actually are, and how image transformations happen. - [Storage uploads](https://docs.powabase.ai/guides/storage-uploads.md): Upload files to Powabase Storage: simple multipart, browser-direct with the Anon Key, server-side signed upload URLs, and TUS resumable uploads for large files. - [Storage policies](https://docs.powabase.ai/guides/storage-policies.md): Three copy-paste RLS patterns for storage.objects: own-files-only, public-read with auth-write, role-based access. Plus the storage.foldername helper and the pitfalls that bite people. - [Storage](https://docs.powabase.ai/api-reference/storage.md): End-user-facing Storage endpoints at /storage/v1/* — bucket CRUD, object upload/download/list/copy/move, signed URLs, image transformations, and TUS resumable uploads. - [Realtime model](https://docs.powabase.ai/concepts/realtime.md): Three channel types: Broadcast for ephemeral messages, Presence for online-state sync, Postgres Changes for table change streams. How auth, RLS, and the per-project tenancy fit together. - [Realtime subscriptions](https://docs.powabase.ai/guides/realtime-subscriptions.md): Three worked patterns: a live-updating list driven by postgres_changes, a presence-based 'who's online' indicator, and a Broadcast chat. - [Realtime](https://docs.powabase.ai/api-reference/realtime.md): WebSocket protocol for subscribing to Broadcast, Presence, and Postgres Changes events, plus the REST endpoint for server-side message emission. - [Billing model](https://docs.powabase.ai/concepts/billing-model.md): How Powabase charges in credits, what triggers a charge, the 402 and 503 errors users see when they hit their balance or billing is unreachable, and where BYOK keys fit in. - [Rate limits](https://docs.powabase.ai/concepts/rate-limits.md): Where Powabase enforces request-rate limits today (workflow executions at 20/min) and where it deliberately doesn't (the rest of the surface). - [Observability](https://docs.powabase.ai/concepts/observability.md): What's user-callable for monitoring your project's health, and what stays internal to the platform infrastructure. Honest framing of where the line is drawn today. - [Backups and disaster recovery](https://docs.powabase.ai/concepts/backups-and-dr.md): What the platform automatically backs up, what's NOT user-callable, and what to do if you need a restore. Honest about the gap between 'we run backups' and 'you can restore yourself.' - [BaaS + AI cookbook](https://docs.powabase.ai/guides/baas-ai-cookbook.md): Four worked patterns that pair the BaaS primitives (PostgREST, Auth, Storage, Realtime) with the AI surface (agents, KBs, workflows). The composability story Powabase is built around. - [Authentication & Storage](https://docs.powabase.ai/api-reference/auth-storage.md): Auth user management and storage operations are served by the control-plane proxy, not the per-project service API. Routing is ref-only: /api/platform/auth/{ref}/* and /api/platform/storage/{ref}/*. - [API conventions](https://docs.powabase.ai/concepts/api-conventions.md): The shared patterns across Powabase's HTTP APIs: header conventions, naming, error envelopes, pagination, response shapes. Once you know these, every endpoint is easier to read. - [API access (raw HTTP)](https://docs.powabase.ai/concepts/api-access.md): Powabase ships no typed SDKs today. Raw HTTP against the documented endpoints is the supported access path. What that means for your stack, and what's likely coming. - [Glossary](https://docs.powabase.ai/concepts/glossary.md): Disambiguating the terms Powabase uses that overlap with each other or with terms from other platforms: agent sessions vs auth sessions, agent hooks vs DB webhooks, etc. - [Common pitfalls](https://docs.powabase.ai/concepts/common-pitfalls.md): The footguns Powabase users hit most: silent failures, wrong field placements, the backend-only ai schema, the differences between agentic and database webhooks. With concrete fixes. - [Migrating from Supabase](https://docs.powabase.ai/guides/migrating-from-supabase.md): What's identical, what's different, what breaks. Database URL format, pooler choice, the absent extensions, the ai schema as new surface, and the agentic API as the headline addition. - [Self-host (enterprise)](https://docs.powabase.ai/guides/self-host-enterprise.md): Self-hosting Powabase is an enterprise offering: what gets deployed, how it differs from managed cloud, and how to engage. Not a publicly self-serve product. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.