Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Show HN: Budibase Agents Beta – model-agnostic AI agents for internal workflows (budibase.com)
7 points by mjashanks 32 days ago | hide | past | favorite | 1 comment
Hi there, Mike here from Budibase.

We’ve recently launched AI Agents into Beta, alongside our existing open-source app-building and automation tools.

We built Budibase Agents for teams that want to leverage AI within real-world workflows, using their own LLMs, data, and APIs. Because of this, our Agents can be powered by any LLM with an OpenAI-compatible API, including open-source and locally-hosted models. This means you can build Agents that connect to your existing toolstack, within your own environment.

Some more details:

- Agents' behavior is configured using natural language instructions, within existing Budibase Workspaces.

- You explicitly control which data sources, APIs, and automations an agent can access.

- End-users can interact with your agents via Budibase Chat, or using existing chat tools like Slack and Discord.

- Agents can be called from Automations, and vice versa, enabling complex workflows, including interacting with end-user apps for manual human approvals.

The Budibase Agents Beta is available now for all self-hosted and cloud users.

As this is a Beta release, we’re actively looking for input and feedback on how we can improve Agents, in terms of experience and functionality for building using AI within real-life workflows.

Feel free to let us know via our GitHub Discussions: https://github.com/budibase/budibase/discussions



[flagged]


At the moment any tool that you write inline in your instructions - the agent has access to use.

The way to scope down access, would be to use an Automation (our RPA, zapier-like section). You can give an automation parameters and control exactly what it does, then use it as an agent tool.

In future, we'll be looking into user-delegate permissions. I.e. you give your users a role (RBAC) in Budibase, which gives them access to certain resources. Any agent calls will be done under the user's permissions. This is not always desirable - sometimes you want your agent to have more permissions that the user - so this will need to be configurable at some level.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: