Give your agent the next task without waiting for it to finish. You can now queue messages in Agenta while your agent works. For example, while it researches a company, queue an instruction to draft an outreach email using that research. You can edit a queued message before it starts, so you can change the brief while the current task continues.
Agenta
Technology, Information and Internet
The open-source workspace for building and running AI agents. Build agents through chat, share them with your team.
About us
Agenta is an open-source workspace where teams create AI agents for engineering, customer support, sales, company knowledge, and operations. Employees can work with these agents in chat and connect them to company files and apps. Teams can share the agents, improve them with feedback, or have them run automatically on a schedule or when an event happens.
- Website
-
https://www.agenta.ai
External link for Agenta
- Industry
- Technology, Information and Internet
- Company size
- 2-10 employees
- Headquarters
- Berlin
- Type
- Privately Held
- Founded
- 2023
Employees at Agenta
View 11 employees at Agenta
or
Locations
-
Primary
Get directions
Berlin, DE
Updates
-
New in Agenta v0.114.3 You can put the keyboard aside and talk to your agent in Agenta. Hit the microphone button, say what you need, and it turns your voice into text. You can edit it before sending. On desktop, you can also hold Ctrl+Option on Mac or Ctrl+left Alt on Windows and Linux. Release to stop talking, then send when you're ready.
-
You can now give each agent in Agenta its own icon and color. Click the icon next to the agent's name and pick one. You'll see it in the sidebar, chat, and automation picker. The choice is saved in your current browser. https://agenta.ai
-
Build an AI agent that helps you prepare for sales calls. Our new walkthrough follows the setup in Agenta, from connecting tools to creating an automation. You’ll see how to: • Connect Granola and ask questions about previous calls. • Give the agent a role, business context, and rules. • Work together on a prospect brief and improve it with feedback. • Save the process as a skill. • Ask the agent to prepare and email a brief when a new sales call appears in your calendar. The video shows the prompts and the steps in the product, so you can follow along with your own use case. Tutorial in the comments
-
Agenta reposted this
Here is a simple trick that makes your AI agents much more useful: Stop thinking of skills as something fixed. Instead each time you use them: 1. Ask the agent to reflect and propose improvements to the skill so that the work next time is easier 2. Give the agent explicit feedback and ask it to incorporate it in the skill Just like a human-intern, the agent will get better at the task and become more useful.
-
Agenta reposted this
Most people automate with AI in the wrong way (then think AI is overhyped). Before you automate a task with an AI agent, you need to see it do the job well at least once (just like an intern). In this walkthrough, the job is preparing for sales calls: research the prospect and send me a brief before the call. I work through an example with the agent, review the result, and give feedback on what is missing. Then I ask it to save those instructions as a skill. Once I’m happy with the result, I ask it to run that skill when a new sales call appears in my calendar. Here is the concrete way to build an AI coworker: • Give it a role, context, and access to the right tools. • Solve a real task together. • Save what worked as a skill. • Automate it, then keep reviewing the results. I recorded the full process in @agenta, including the prompts and the changes along the way. Watch the full walkthrough in the link in the comments 👇️
-
Agenta is the work of 100+ contributors and we're looking for more! You don't need to code to contribute, you can work on docs, tutorials, design and many things. Join our Slack if you're interested (link below)!
🚀 My first open-source contribution has been merged! I'm excited to share that my first contribution to Agenta has been merged into the project. 🐛 The issue #5661: Agenta's execution layer treats pi_core as the default harness when no harness is explicitly specified. However, the Entity UI was treating a missing harness as null. Because of this mismatch, Pi-specific configuration controls — including the Permissions section — were hidden when the harness wasn't explicitly present in the configuration. 🔧 The fix: I aligned the Entity UI with the execution layer by resolving a missing harness to pi_core and added regression tests covering missing, explicit, and invalid harness values. Through this contribution, I got hands-on experience with: 🔍 Debugging a real-world codebase 🧪 Writing regression tests with Vitest 🛠️ TypeScript/React 🔀 Git & GitHub pull-request workflow 👨💻 Working through maintainer code review What made this especially meaningful was seeing my GitHub username appear in Agenta's contributors section after the contribution was merged. ❤️ Huge thanks to the Agenta maintainers for the review, guidance, and opportunity to contribute. #OpenSource #GitHub #SoftwareEngineering #TypeScript #React #WebDevelopment
-
