Multi-Agent Builder - Secure agents that collaborate and take action across tools
Credal enables enterprises to deploy secure, collaborative AI Agents to handle multi-step workflows across an organization’s tools, data and subject matter expertise. Agents can autonomously call other agents and take action while respecting data permissions.
Replies
Credal.ai
Looks like services connection middle-ware, nice idea!
Great for secure workflows; I appreciate the focus on compliance! 👍
Credal.ai
@shenjun Thanks Jun! Has been our bread and butter from day one- its just what it takes to get this deployed in the contexts that actually matter the most to the world
Fable Wizard
Credal’s multi-agent platform is such a cool solution for enabling seamless collaboration across tools while keeping data secure and compliant—this could really streamline operations in enterprises! What’s been the most surprising use case you’ve seen so far that you didn’t initially expect?
Credal.ai
@jonurbonas Thanks Jonas!
There's been so many surprising usecases!
- A global regulator/standards setter called IFRS uses Credal Agents to set global accounting standards! When they issue new draft standards, they'll use on Agent to comb through all the comment letters they get from accountants, governments and companies and synthesize the key points of feedback from each group. That agent can then invoke a separate writer agent to draft clarification letters / statements to address the points of feedback raised, which a human can review and publish. DIdn't even know this process existed before Credal!
- Some simpler, but perhaps more generalizable usecases includes getting one Agent to read through github and slack to identify changes, then combining that with another to draft marketing content and setup the linkedin post!
Kona
Congrats on the launch! With agents calling agents, curious how you all think about alignment w/ user intent?
Credal.ai
@andrewthezhou Thanks Andrew! Alignment with user intent is actually really tied up with how we think about security and governance. Its all about providing the user full visibility into how the AI is behaving (and why), and the ability to control that behaviour, by tweaking things like the prompting, the training examples, the search strategy, etc
Looks really cool!
Question - do you guys support function calling? Like if I need to trigger some actions in response to user input, for example
All the best for the launch @ravin_thambapillai & team!
Credal.ai
@parekh_tanmay Thanks for the support!
Flex-Worthy Templates
I think businesses would love to use this tool to have agents take actions across tools easily
Credal.ai
@shushantlakhyani thanks!
Love the ‘agents calling agents’ concept. How do you prevent recursive loops or conflicting actions? Is there a built-in protocol for resolving disputes between agents with competing priorities?
Credal.ai
@desmond_ren1 great questions!! For now we just have a max depth limit (which is configurable but starts at 5). Sometimes you do actually want loops so we don’t ban that (E.g you have a content writer and a content critiques, so the writer gets the critique, improves the writing, then gets more critiques etc until the feedback is no longer improving the content).
For agents that disagree, we typically recommend people create their agent teams with a single orchestrator or decision maker at the top that then arbitrates between sub agents!
This is sick, love the implementation of multi agents here. I've run into issues with llms getting overwhelmed with the amount of functions I throw at them, so grouping these actions together in independent systems makes a lot of sense to me. Visualization is pretty neat as well 👀
Open Census Data
Wow, this is really cool – def something I can use across our AI powered workflows at DataLemur. Congrats Richad & the whole Credal team on the launch!
ThreeDee
Collaborative AI agents for workflows sound like a productivity boost! 👌
Billy
What's the pricing model?
AI Agent is the future
Credal is an innovative and powerful solution for enterprises. Its ability to deploy secure, collaborative AI Agents that can handle complex workflows across various tools and data sources is a game-changer. The autonomy of agents to call other agents while respecting data permissions adds an extra layer of security and efficiency. This tool has the potential to significantly enhance productivity and streamline operations in organizations of all sizes.
Late to the party here but curious if data permissions/GRC policies would be highly customized per enterprise? Thanks!