Raven is an open-source team chat application built on the Frappe Framework, the stack ERPNext runs on, and it lives inside your ERPNext instance instead of a separate Slack or Teams subscription. Since Raven v2 it also has no-code AI agents that can read and create ERPNext records and pull data out of files such as receipt images, using OpenAI or a self-hosted model.
Raven started as an answer to a common gap: ERPNext teams needed internal chat, and buying Slack meant paying for and maintaining yet another tool alongside the ERP. Raven closes that gap inside the same Frappe system: channels, direct messages, threads, and alerts tied to ERPNext records. It began at The Commit Company; its repository now sits under frappe/raven, licensed AGPL-3.0, and v3.0.0 shipped in September 2026 with a faster web app, an improved PWA, scheduled messages and reminders.
What makes Raven worth a closer look is the AI agent layer, which turns it from an internal Slack alternative into something that can act on your ERPNext data through conversation. I lead ERPNext and Frappe programmes at MithTech, a team of Frappe specialists in Bengaluru, and this is what Raven does today.
What does Raven do inside ERPNext?
Raven provides team chat (channels, direct messages, threads and document-linked notifications) running inside an ERPNext instance. You can link and preview ERPNext documents in chat, run workflow actions and print documents from a conversation. Since v2 it also has no-code AI agents that read an instruction in a chat message and act on ERPNext data in response. The chat layer alone can replace Slack or Teams for internal communication; the agent layer is newer and less widely understood.
Document notifications deserve their own mention. Raven posts to a channel when a specific ERPNext document event happens: a Sales Order submitted, a Purchase Invoice approved. The conversation stays tied to what is actually happening in the ERP, instead of a separate activity feed nobody checks.
Skimmable summary: Raven is team chat inside ERPNext, with channels, DMs and document-linked notifications, plus no-code AI agents that act on ERPNext data from a conversation. Two separate capabilities in one app.
How do Raven's AI agents work?
You define an agent in Raven's interface: its instructions, the model it uses, and the functions it may call, such as getting a document, creating one, or a custom function like fetching a supplier by GST number. In response to a chat message, the agent can call those functions to read, create or update ERPNext records, and it can extract data from an uploaded file, such as a receipt or invoice image, to draft the matching document. No custom code is needed for the standard functions.
Raven v2 launched on OpenAI's Assistants API. Since v2.6 (mid-2025), agents run on the OpenAI Agents SDK, which also let Raven add local models. OpenAI has been moving developers from the Assistants API to its Responses API, so if you set up agents on an older Raven version, upgrade before relying on them.
Permissions follow the user, not the agent. An agent can only call the functions you have given it, and those functions run Frappe's normal permission checks for the person using it in that conversation. If a user cannot create a document type, the agent acting for them cannot either, so an agent's access never silently exceeds the human's.
Skimmable summary: Raven agents are configured without code, call only the functions you define, can read files such as receipts, and inherit the permissions of whoever uses them.
Can Raven run without sending data to OpenAI?
Yes. Raven supports local models through Ollama and other OpenAI-compatible model servers that you host yourself, on your own hardware or a cloud server you control. For teams where data privacy is a hard requirement, this is the option that matters: the agent runs against a model you host rather than a third-party API.
The trade-off is real. Open models you can run through Ollama are generally less capable than the newest cloud models from OpenAI or Anthropic. Running locally buys privacy at some cost in capability; it is not a free substitute. Test against your actual use cases before committing either way.
How do I set up Raven on my ERPNext instance?
Install Raven as a Frappe app on your existing bench, or from the Apps tab on Frappe Cloud. Channels and direct messages work straight away. The AI layer is a separate step; do it in this order:
- Install the app.
bench get-app raven && bench --site <site> install-app ravenon a self-hosted bench, or the Apps tab on Frappe Cloud. - Check chat works first. Create a channel and send a message. If this fails, no model configuration will help.
- Set up only the document notifications you want. Start with one or two, such as a submitted Sales Order or an approved Purchase Invoice. Alerting on everything trains people to ignore the channel.
- Connect a model. An OpenAI API key, or Ollama (or another OpenAI-compatible server) pointed at a local model.
- Define one agent, scoped narrowly. Give it instructions and only the functions and DocTypes it needs, read-only to begin with.
- Test it on staging. Have it draft documents you check by hand before it gets write access.
- Grant write access last, one DocType at a time, once its drafts have been right for a while.
On step 4, the model choice is a real trade-off:
| Where your data goes | To OpenAI's API | Stays on infrastructure you control |
| Capability | Current OpenAI models | Generally behind the newest cloud models |
| Running cost | Per-token API billing | Your own compute, no per-token fee |
| Hardware needed | None | A machine that can hold the model |
| Setup effort | An API key | Install, host and maintain the runtime |
| Fit | Capability matters more than residency | Data residency or policy rules out a cloud LLM |
Frequently asked questions
+Is Raven free to use with ERPNext?
Yes. Raven is open source under AGPL-3.0, with no licence fee, like the rest of the Frappe ecosystem. If you connect a cloud model like OpenAI for the AI agents, you pay that provider for usage; Raven itself adds no cost.
+Do I need coding experience to set up Raven's AI agents?
No. Agents are configured in Raven's interface, not built as a custom Frappe app. Connecting a model and choosing the functions and DocTypes an agent may use is configuration work. Custom functions beyond the standard ones need a developer.
+Can Raven agents access data I don't have permission to see?
No. Agents can only call the functions you define, and those run Frappe's normal permission checks for the user in the conversation. If that user cannot read or create a document, the agent acting for them cannot either, however it is instructed.
+Is Raven the same as ERPNext's other AI options, like Frappe Assistant Core?
No, they solve different problems. Raven is chat-native, with AI agents that act inside a conversation. Frappe Assistant Core connects ERPNext to Claude or ChatGPT over the Model Context Protocol (MCP) for natural-language querying and reporting outside a chat app. Many ERPNext AI setups use both.
Comments & ratings
No comments yet. Start a new discussion.
Manoj · Published on 17 July 2026
Worked with us? Leave a Google review