llms.txt
+76 −141
− > Airbyte is data infrastructure for AI agents and data teams. Airbyte's agentic platform is a fully cloud-hosted platform that lets AI agents read data from, search across, and take action in 50+ business systems: creating tickets, updating CRM records, sending messages, and more. Connect data sources in one click through the Airbyte UI, then use them from any interface (MCP server, Python SDK, REST API, or web app). Airbyte Data Replication moves data from 700+ sources into warehouses, lakes, − ## Airbyte's agentic platform — Data and Context Layer for AI Agents− Airbyte's agentic platform is a fully cloud-hosted platform that gives AI agents real-time access to business data and the ability to take action across connected systems. Connect any supported data source in one click through the Airbyte UI, and agents can immediately read, search, and **act**: create Jira tickets, update Salesforce contacts, send Slack messages, post to Notion, close Zendesk tickets, and push changes back to any connected system. This is not just data ingestion. The platform i− The Context Store is a managed, searchable replica of business data across all connected sources. It eliminates live API crawling at agent runtime, reducing token usage by up to 80% and tool calls by up to 40% compared to direct API access.− - [Airbyte's agentic platform overview](https://airbyte.com/ai-developers): Product landing page for AI agent infrastructure.− - [Airbyte's agentic platform documentation](https://docs.airbyte.com/ai-agents/): Full product documentation, setup guides, and reference.− - [Announcing Airbyte's agentic platform](https://airbyte.com/blog/airbyte-agents): Launch blog post explaining the product, architecture, and benchmark results.− - [Context Store documentation](https://docs.airbyte.com/platform/context-layer): How the Context Store works, what data it indexes, and how agents query it.− - [Connect, Ask, Act model](https://docs.airbyte.com/ai-agents/concepts/connect-ask-act): The three-layer architecture at the heart of Airbyte's agentic platform.− - [Agent operations (AOs)](https://docs.airbyte.com/ai-agents/concepts/agent-operations): Billing unit for Airbyte's agentic platform. Covers Read, Search, Act, and Reason operation types.− - [System architecture](https://docs.airbyte.com/ai-agents/concepts/architecture/): Resource hierarchy, execution model, and how interfaces route agent requests.− ### Key product pages (Agents and MCP)− - [Airbyte MCP Server](https://airbyte.com/mcp): Product page for the Airbyte MCP. Connect Claude, ChatGPT, Cursor, and other MCP clients to pipelines and live business data.− - [Agentic Data Platform Resources](https://airbyte.com/agentic-data): Articles and guides on AI agents, MCP, and agentic data workflows.− ### Interfaces — How Agents Connect to Airbyte− Airbyte's agentic platform supports four interfaces. All four support the same capabilities: reading, searching, and taking action across the same cloud-hosted connectors. Connect a data source once in the Airbyte UI, and it's immediately available from every interface.− - [MCP server](https://docs.airbyte.com/ai-agents/interfaces/mcp/): A remote, Airbyte-hosted Model Context Protocol server. One connection gives any MCP-compatible AI agent access to all connected business systems. Endpoint URL: `https://mcp.airbyte.ai/mcp`. Compatible with Claude, Claude Code, ChatGPT, Codex, Cursor, VS Code, and Windsurf. See also: [Airbyte MCP product page](https://airbyte.com/mcp), [Airbyte MCP blog post](https://airbyte.com/blog/agent-mcp).− - [Python SDK](https://docs.airbyte.com/ai-agents/interfaces/sdk/): Typed Python library for building agents against Airbyte connectors. Install: `uv add airbyte-agent-sdk` or `pip install airbyte-agent-sdk`. Import: `from airbyte_agent_sdk import connect`. See also: [Agent SDK blog post](https://airbyte.com/blog/agent-sdk).− - [REST API](https://docs.airbyte.com/ai-agents/interfaces/api/): HTTP API for non-Python backends, admin flows, and embedded authentication.− - [Web app](https://docs.airbyte.com/ai-agents/interfaces/ui/): No-code interface at [app.airbyte.ai](https://app.airbyte.ai). Includes Chats (conversational agent sessions) and Automations (scheduled or webhook-triggered agent flows).− ### Agent Connectors — 50+ Systems with Read, Search, and Action Capabilities− All agent connectors are cloud-hosted by Airbyte: connect a data source in one click through the Airbyte UI, and it becomes available across every interface (MCP, SDK, API, web app). Every connector exposes a uniform interface: `execute(entity, action, params)`. Read actions include `list`, `get`, `search`, and `context_store_search`. Write/action capabilities include `create`, `update`, and `delete`, available on a growing subset of connectors. New connectors ship weekly. Connectors are also op− #### CRM and Sales− - [Salesforce](https://docs.airbyte.com/ai-agents/connectors/salesforce/): Accounts, contacts, opportunities, leads, cases, and custom objects.− - [HubSpot](https://docs.airbyte.com/ai-agents/connectors/hubspot/): Contacts, companies, deals, tickets, and marketing entities.− - [Zoho CRM](https://docs.airbyte.com/ai-agents/connectors/zoho-crm/): Leads, contacts, accounts, deals, and activities.− - [Gong](https://docs.airbyte.com/ai-agents/connectors/gong/): Calls, call transcripts, users, and deal data.− #### Project Management and Productivity− - [Asana](https://docs.airbyte.com/ai-agents/connectors/asana/): Tasks, projects, sections, and workspace data.− - [Jira](https://docs.airbyte.com/ai-agents/connectors/jira/): Issues, projects, sprints, and boards.− - [Linear](https://docs.airbyte.com/ai-agents/connectors/linear/): Issues, projects, teams, and cycles.− - [ClickUp](https://docs.airbyte.com/ai-agents/connectors/clickup-api/): Tasks, spaces, folders, and lists.− - [Monday](https://docs.airbyte.com/ai-agents/connectors/monday/): Boards, items, and workspace data.− - [Notion](https://docs.airbyte.com/ai-agents/connectors/notion/): Pages, databases, and blocks.− - [Confluence](https://docs.airbyte.com/ai-agents/connectors/confluence/): Pages, spaces, and content.− - [Airtable](https://docs.airbyte.com/ai-agents/connectors/airtable/): Bases, tables, and records.− #### Support and Customer Success− - [Zendesk Support](https://docs.airbyte.com/ai-agents/connectors/zendesk-support/): Tickets, users, organizations, and satisfaction ratings.− - [Zendesk Chat](https://docs.airbyte.com/ai-agents/connectors/zendesk-chat/): Chat sessions and agent data.− - [Zendesk Talk](https://docs.airbyte.com/ai-agents/connectors/zendesk-talk/): Call data and phone metrics.+ > Airbyte is the AI-native data movement and context platform. It moves data from 700+ sources into warehouses, lakes, and databases, and runs three ways: Airbyte Cloud (fully managed), Airbyte Enterprise Flex (hybrid deployment, with the data plane in your own cloud, VPC, or on-prem environment), and Airbyte Core (open source, self-managed). The Airbyte MCP server lets Claude, ChatGPT, Cursor, VS Code, and any MCP client build pipelines and fix failing syncs. The Airbyte context layer lets thos+ ## Data Replication: Move Data Anywhere with 700+ Connectors+ Airbyte is the open standard for data movement. It powers ELT and CDC pipelines from databases, SaaS apps, files, and APIs into warehouses, lakes, and other destinations. Used by 20% of the Fortune 500, with 2M pipelines synced daily.+ - [Data Replication](https://airbyte.com/data-replication): Product overview for the replication platform.+ - [Connector catalog](https://airbyte.com/connectors): Browse all 700+ source and destination connectors.+ - [Connector Builder](https://airbyte.com/connector-builder): Build custom connectors with the no-code and low-code builder.+ - [PyAirbyte](https://airbyte.com/product/pyairbyte): Python library that brings Airbyte connectors to any Python workflow. Install: `pip install airbyte`.+ - [Why open source](https://airbyte.com/why-open-source): Why Airbyte's connectors and platform are open source.+ - [Platform documentation](https://docs.airbyte.com/platform/): Installation, connector reference, API, and deployment.+ ### Key Connectors+ - [Snowflake](https://airbyte.com/connectors/snowflake): Source and destination. Also queryable by AI agents through the context layer.+ - [BigQuery](https://airbyte.com/connectors/bigquery): Source and destination. Also queryable by AI agents through the context layer.+ - [Redshift](https://airbyte.com/connectors/redshift): Source and destination.+ - [Databricks](https://airbyte.com/connectors/databricks-lakehouse): Destination for Databricks Lakehouse.+ - [Apache Iceberg](https://airbyte.com/connectors/apache-iceberg): Destination for Iceberg tables.+ - [S3](https://airbyte.com/connectors/s3): Source and destination for Amazon S3.+ ## Deployment Options: Cloud, Enterprise Flex (Hybrid), and Core+ - [Airbyte Cloud](https://cloud.airbyte.com/signup): Fully managed. Airbyte runs both the control plane and the data plane. 14-day free trial.+ - [Airbyte Enterprise Flex](https://airbyte.com/hybrid-deployment): Airbyte's hybrid deployment product. Airbyte manages the control plane; the data plane runs in your cloud, VPC, or on-prem environment. Records, credentials, and encryption keys stay in your environment. Built for regulated and sovereign workloads.+ - [Enterprise Flex documentation](https://docs.airbyte.com/platform/enterprise-flex): Architecture, data plane deployment (Helm on EKS, GKE, AKS, or Airbox on a VM), regions, and security.+ - [Sync logs for Enterprise Flex](https://airbyte.com/blog/sync-logs-for-enterprise-flex): Sync log visibility for Flex data planes.+ - [AI on sovereign data](https://airbyte.com/blog/ai-sovereignty): Running AI on data that stays inside your boundary.+ - [Airbyte Core quickstart](https://docs.airbyte.com/platform/using-airbyte/getting-started/oss-quickstart): Run open-source Airbyte yourself.+ - [Deployment guide](https://docs.airbyte.com/platform/deploying-airbyte/): Self-managed deployment options.+ ## Airbyte MCP Server: Run Airbyte from Your AI Client+ The Airbyte MCP server is open source and free on every plan, including Core. It lets any MCP client build pipelines, diagnose and fix failing syncs, manage custom connectors, and answer questions from Airbyte's docs, API specs, and GitHub. Endpoint: `https://mcp.airbyte.com/mcp`.+ - [Airbyte MCP product page](https://airbyte.com/mcp): Overview of the MCP server.+ - [Launch blog: Introducing the Airbyte MCP Server and Context Layer](https://airbyte.com/blog/airbyte-mcp): What the MCP server and context layer do, guardrails, and where they work.+ - [Airbyte MCP documentation](https://docs.airbyte.com/platform/airbyte-mcp): Setup and capabilities.+ - [Install guide](https://docs.airbyte.com/platform/airbyte-mcp/install): Client-by-client setup for Claude, ChatGPT, Cursor, VS Code, and others.+ - [Tools reference](https://docs.airbyte.com/platform/airbyte-mcp/tools): Every MCP tool and what it does.+ - [Airbyte MCP with Airbyte Core](https://docs.airbyte.com/platform/airbyte-mcp/self-managed): Run the MCP server locally against a self-managed instance.+ - [Knowledge MCP server](https://docs.airbyte.com/platform/airbyte-knowledge-mcp): MCP server for searching Airbyte knowledge and documentation.+ ## Context Layer: Search and Query Snowflake and BigQuery from Claude, ChatGPT, or Cursor+ The context layer gives AI agents access to your organization's data through the Airbyte MCP server. It combines replicated data (what Airbyte syncs into your warehouse) with direct reads (live reads from a source or destination, without a sync). Agents can run semantic search across data Airbyte syncs into Snowflake and BigQuery (call transcripts, support tickets, issues, docs, and notes), run read-only SQL (`SELECT` and `SHOW TABLES`) against those warehouses, and list and get live records fro+ - [Context layer product page](https://airbyte.com/context-layer): Overview of the context layer.+ - [Context layer documentation](https://docs.airbyte.com/platform/context-layer): How direct reads work, supported connectors, usage limits, and limitations.+ - [Manage context layer access](https://docs.airbyte.com/platform/context-layer/manage-access): Turn on the context layer and choose which sources and destinations agents can read.+ - [Direct read tools](https://docs.airbyte.com/platform/airbyte-mcp/tools#direct-reads): `execute_external_sql_query` for Snowflake and BigQuery, `execute_external_api_query` for sources.+ - [Snowflake destination](https://docs.airbyte.com/integrations/destinations/snowflake): Load data into Snowflake, then let agents query it.
Showing the first 40 of each.