Agent Cloud
Agent Cloud is an open-source platform designed to enable companies to build, deploy, and engage with private LLM (Large Language Model) chat applications. Thes...
Last verified:
What is Agent Cloud?
Agent Cloud is an open-source platform that enables companies to build and deploy private LLM chat apps (like ChatGPT) on their own infrastructure, allowing teams to securely interact with their data. The platform serves as a self-hosted GPT builder with additional capabilities, offering two main types of applications: Conversational Chat Apps for single-agent chat similar to OpenAI GPTs, and Process Apps for multi-agent process automation that allocates goals and tasks for agents to complete.
Key features include a built-in RAG pipeline that natively embeds 260+ data sources out of the box, enabling users to split, chunk, and embed data from sources like Confluence, databases (Postgres, Snowflake, BigQuery), and file uploads (PDF, DOCX, TXT, CSV, XLSX). The platform is LLM-agnostic, supporting any LLM with an OpenAI-compatible endpoint including Ollama, LM Studio, OpenAI, and Azure OpenAI. It includes vector database support (Qdrant, Pinecone), data sync frequency controls (manual, scheduled, cron expressions), multi-agent engine powered by crewai, and comprehensive team/user permissions systems.
Agent Cloud is designed for developers, engineers, and companies who want to deploy secure AI applications for their teams including sales, HR, operations, customer support, and education sectors. The platform serves organizations with strict regulatory requirements or sensitive data handling policies who need fully local deployment options.
The platform supports both managed cloud version and self-hosted open source deployment, with AGPL 3.0 licensing for the open source version. It includes ELT functionality through Airbyte integration, semantic chunking capabilities, and API exposure for vector upsert operations.
Agent Cloud pricing
Pricing model: Free
Agent Cloud offers a freemium pricing model with both free and paid accounts. The open source version is available under AGPL 3.0 license for self-hosting on your own infrastructure with no cost. The managed cloud version has free and paid tiers - check their pricing page for specific plan details. Free tier includes community edition access. Paid plans vary by features including sync frequency controls and cloud infrastructure. The platform uses BYO token model where users bring their own LLM API keys (OpenAI, Azure OpenAI) rather than paying for tokens through Agent Cloud.
Agent Cloud pros
- Open source with AGPL 3.0 license for self-hosting
- 260+ native data source connectors out of the box
- LLM agnostic - supports any OpenAI-compatible endpoint
- Full local deployment option for maximum data privacy
- Built-in end-to-end RAG pipeline with vector DB
- Supports both single-agent chat and multi-agent process apps
- Integrates Airbyte for ELT and Qdrant for vector storage
- Semantic chunking for smarter data splitting
- Flexible sync frequency (manual, hourly, daily, cron)
- Team and user permission management system
- Supports local embedding models via fastembed
- Multiple file upload formats (PDF, DOCX, TXT, CSV, XLSX)
- Database column/table selection for targeted ingestion
- CrewAI-powered multi-agent runtime abstraction
- Open source with active Discord community for support
- BYO token model - you bring your own LLM API keys
Agent Cloud cons
- Requires 16 GB RAM minimum for Docker installation
- 8GB Macbook Air M1/M2 cannot run Agent Cloud effectively
- Windows users must use WSL for installation
- Only Mac/Linux have direct docker install.sh script
- Cloud inference currently limited to OpenAI and Azure OpenAI
- Self-hosting requires additional RAM for local LLM inference
- Longer setup complexity compared to fully managed solutions
- Cloud sync frequency depends on paid plan limitations
Frequently asked questions about Agent Cloud
What is Agent Cloud?
Agent Cloud is a single place where companies can build and deploy AI apps. This includes single agent chat apps, multi agent chat apps, and knowledge retrieval apps. The platform enables developers and engineers to build apps for themselves and teams they interface to in sales, HR, operations, and other departments. It functions as a self-hosted GPT builder platform with extra goodies including RAG pipeline, multi-agent engine, and 260+ data source connectors.
What software license does Agent Cloud use?
Agent Cloud uses AGPL 3.0, which is a copy-left license. The license can be found on their GitHub page. This open source license allows companies to self-host the platform on their own infrastructure while maintaining the ability to use it privately for their employees.
What hardware requirements are needed to run Agent Cloud locally?
For Docker installation, you strongly need a machine with at least 16 GB of RAM. A base Macbook Air M1/M2 with 8GB RAM will not suffice as Airbyte requires more resources. If running Ollama or LM Studio locally for local LLM inference, you'll need additional RAM. Without Docker, 8GB RAM may suffice but it is harder to get started.
Can I use a local Large Language Model with Agent Cloud?
Yes. Agent Cloud supports any local LLM which has an OpenAI compatible endpoint. This means you can use LM Studio or Ollama with the app for local inference. For cloud inference, they currently support OpenAI and Azure OpenAI, with long-term vision to support all cloud providers using the litellm library.
How many data sources does Agent Cloud connect to?
Agent Cloud comes with a built-in data pipeline that allows you to split, chunk, and embed data from over 260 sources out of the box. This includes connectors for systems like Confluence, databases (Postgres, Snowflake, BigQuery), and supports file uploads including PDF, DOCX, TXT, CSV, and XLSX formats.
What vector databases does Agent Cloud support?
For self-hosting, Agent Cloud connects to major Vector DBs including Qdrant and Pinecone. The default vector DB is Qdrant, located at http://localhost:6333/dashboard#/collections for local access. If signed up for the cloud version, you can use RAG with synced data directly by building chat apps or using APIs.
Can I control sync frequency for data sources?
Yes, you can select basic or advanced sync frequency options, giving you hourly, daily, or cron expression syncs. For cloud deployments, sync frequency depends on your plan. To edit, go to Data Sources > Select Data Source > Select Schedule Tab and click [edit schedule].
How can my data retrieval be truly private?
For truly private data retrieval, do not use the managed cloud product or anyone else's cloud service. The recommendation is to deploy the open source app on your own infrastructure, run an LLM on-prem or in your cloud, and sign an enterprise self-managed license. This ensures maximum control over data privacy and sovereignty.
What splitting and chunking methods does Agent Cloud support?
For files, Agent Cloud supports two methods: Basic character splitting (e.g. /n) and Advanced semantic chunking which leverages an embedding model to chunk semantically similar sentences. For data sources other than files, data automatically chunks by message through Rabbit MQ - for example, Bigquery messages equal one row, so it chunks row by row. Long-term they plan to enable API endpoints for custom chunking strategies.
How can I contribute to the Agent Cloud repository?
If you want an initial idea of how your code fits into the repo, raise a feature request first. If you're keen to contribute immediately, write the code and raise a pull request. You can also chat on their Discord server for community support and collaboration with other contributors.