Today, we’re launching Operly Agents into Beta, empowering our users to build custom agents with their own models, APIs, and data. With Agents, we’re solidifying Operly’s position as the all-in-one open-source AI workflow toolkit.
Introducing Operly Agents
Operly Agents enable you to securely connect LLMs to data and tools to create intelligent AI assistants.
Select any model, define agent behavior, and integrate AI into real-world workflows within a flexible, instruction-led interface.
Agents are available today. We’re excited for you to try them out. You can also provide feedback via our GitHub.
The instruction-led agent toolkit
Each Operly Workspace now includes an Agents tab, alongside your existing Data, APIs, Apps, and Automations. Here, we can define instructions for Agents in natural language, including configuring behavior and making tools and data available to our agents.
We can use the Chat preview to test and iterate over Agent behavior. Operly Agents eliminate the boilerplate required to run AI workflows, providing a fast, maintainable experience for configuring agents that you can roll out with trust and confidence.
Chat with agents, wherever work happens
Alongside Agents, we’ve added a pre-built Chat experience, along with a new Chatbox component to Operly Apps. This provides a seamless experience for end-users to interact with agentic workflows, with the same authentication standards and screen-level RBAC we’d find in any other Operly App.
Agents also support deployment via external chat tools, including Slack, Discord, and Teams, so you can leverage custom AI-powered assistants, wherever your team already works.
Connect to tools and data, securely
Operly Agents can invoke tools and data sources from across your Workspace, including tables, API requests, integrated apps, knowledge sources, and Automations. Agents can only utilize the tools and data we explicitly grant them permissions for.
Operly’s REST templates and built-in database actions make it quick and easy to configure agent tool calls. Simply select which tools your Agents can access, and define how they’re used within your instruction prompt.
Model agnostic, privacy-first
In security-focused teams, AI adoption is often blocked because mission-critical data can’t be shared with opaque public models and services, or even leave their internal environment.
Operly Agents are fully model-agnostic, supporting any LLM with an OpenAI-compatible API, including locally hosted models. Select the right model for the task at hand, while retaining control over data residency within our self-hostable platform.
Build Agents alongside Apps and Automations
Agents are callable within Operly Automations. We can pass data from previous Automation steps and triggers within the instructions field. We can also define output data schemas for downstream processing or utilizing responses in user-facing tools for human-in-the-loop workflows.
Since Operly Automations can be invoked as tools by Agents, this approach enables us to build multi-agent workflows with high levels of control.
Start building agentic workflows today
Our Agents Beta is an important step on Operly’s journey to becoming the all-in-one open-source AI workflow toolkit.
We can’t wait to hear what you think.