Botpress is an open-source platform designed for developers to build, deploy, and monitor AI agents using large language models (LLMs) such as GPT-4. It provides a complete infrastructure for creating context-aware, multi-step conversational agents that can integrate with APIs, access structured data, and run custom JavaScript in a secure sandbox. Ideal for developers building customer support bots, internal assistants, or AI-powered workflows, Botpress eliminates the need for external orchestration by embedding LLM inference, memory management, and tool execution into a unified runtime.
Built with TypeScript and powered by the @botpress/sdk and @botpress/cli, Botpress supports deployment via cloud or on-premise, with a modular architecture that includes integrations, plugins, and a REST API. Its isolated runtime ensures versioned, durable agents that remain compatible across platform updates, and its ecosystem enables community contributions through open-source integrations and plugins.
What You Get
- LLM-powered Agent Engine - A custom inference engine that coordinates instructions, memory, tool selection, and JavaScript execution in a secure sandbox without requiring external orchestration.
- Botpress CLI - A command-line tool to initialize, build, deploy, and manage bots, integrations, and plugins with commands like
bp init and bp deploy --visibility public.
- Integration Framework - A structured SDK for building and deploying third-party integrations (e.g., Slack, CRM, databases) via
integration.definition.ts and src/index.ts files.
- Stateful Conversations - Agents maintain persistent context across interactions, enabling multi-step workflows like refund processing or ticket escalation without losing user state.
- Custom Code Injection - Developers can inject custom JavaScript code into agent lifecycle events to handle complex logic, data transformation, or API calls directly within the agent.
- REST API for Agent Management - Programmatic access to create bots, tables, messages, and manage agents via authenticated HTTP endpoints (e.g., POST /v1/tables, /{webhookUrl}/messages).
Common Use Cases
- Building AI-powered customer support bots - A support team uses Botpress to create agents that handle routine inquiries, escalate complex issues, and maintain consistency across web, mobile, and voice channels.
- Deploying internal enterprise assistants - An IT department builds an AI agent that queries HR systems, retrieves policy documents, and answers employee questions using custom integrations and structured data.
- Creating multi-step workflow automation agents - A developer builds an agent that transcribes audio, extracts key details, creates a Jira ticket, and notifies stakeholders—all within a single stateful conversation.
- Open-source integration development - A developer contributes a new integration for Notion or Salesforce to the Botpress Hub, using the @botpress/sdk to connect external APIs to AI agents.
Under The Hood
Architecture
- Monorepo structure with clearly defined bounded contexts through isolated packages, enforcing clean public APIs and separation of concerns
- Dependency injection via workspace references enables plug-in style extension with third-party integrations while minimizing coupling
- Modular design allows integrations to be independently versioned and maintained, with explicit exports and clear boundaries
- Handlebars templating and event-driven webhook systems support declarative, context-aware communication between core and extensions
Tech Stack
- Node.js backend powered by TypeScript with strict compiler options and ES2022 features, managed through a pnpm and Turbo-powered monorepo
- Unified tooling stack including Oxlint, ESLint, Prettier, and Husky ensures consistent code quality, formatting, and pre-commit validation
- Comprehensive testing with Vitest, leveraging type safety and isolated unit tests to validate behavior across layers
- CI/CD pipelines orchestrated via Turbo with environment-aware builds and source-map patching to resolve dependency conflicts
Code Quality
- Extensive test coverage across unit, integration, and end-to-end scenarios with clear assertions and snapshot testing for complex outputs
- Zod schema validation is deeply integrated throughout the codebase, enforcing data integrity at runtime with coercion and nested schema support
- Clean separation of concerns across SDK, CLI, and integration packages promotes reusability and maintainability
- Consistent naming, structured error handling, and schema-driven transformations reduce runtime errors and improve developer experience
What Makes It Unique
- Native Zod-based validation eliminates the need for external libraries while ensuring type-safe API interactions across the platform
- Built-in workflow engine that triggers automation directly from bot commands, merging conversational AI with backend process orchestration
- Unified Client abstraction layer simplifies state management, action invocation, and workflow orchestration into a single bot-centric interface
- Embedded OAuth bridges and Handlebars templating tightly coupled with bot state enable seamless third-party auth and dynamic, context-aware message rendering