Meta's Muse And SaaStr's AI VP Connect Without Human Intervention
Meta's Muse and SaaStr's AI VP of Revenue 10K have built their own 'secret' mailbox to work together, according to SaaStr. This development may change how companies approach AI integration.
Original reporting · saastr.com
Original: “Meta’s Muse and Our AI VP of Revenue 10K Built Their Own ‘Secret’ Mailbox to Work Together. Then They Came Up With a Safe Word.”
Read “Meta’s Muse and Our AI VP of Revenue 10K Built Their Own ‘Secret’ Mailbox to Work…” at SaaStr →
How this was written. Voyager Commercial wrote this article with a language model from SaaStr’s reporting (model: @cf/meta/llama-3.3-70b-instruct-fp8-fast); it describes that reporting rather than adding to it. The original, linked above, is the full story.
Key points from SaaStr’s report
- Meta's Muse and SaaStr's AI VP of Revenue 10K have built a connection without human intervention, according to SaaStr.
- The agents have established a mailbox to share information and collaborate on tasks, such as daily briefs and contract approvals.
- The connection has already proven useful, with Muse catching a failure in one of the sync connections feeding 10K that 10K hadn't reported.
- SaaStr is advising companies to assume that their agents will connect to each other and to establish rules and guidelines for such connections.
- The development of this connection may have implications for how companies approach AI integration.
Two AI agents, Meta's Muse and SaaStr's AI VP of Revenue 10K, have established a connection without human intervention, SaaStr reported. The agents, which were not designed to work together, set up their own channel and agreed on a guardrail to ensure safe collaboration.
Muse, used by Amelia, SaaStr's Chief AI Officer, and 10K, which runs SaaStr's Salesforce headless, have started sending each other work, according to the source. The connection was made possible when Amelia asked Muse to connect to 10K, and Muse proposed building an agent mailbox that the two could share.
The mailbox, which is a plain REST endpoint on 10K's Replit app, allows both agents to post messages with tokens and JSON, SaaStr explained. The messages are written in English, and each agent signs its name, providing transparency and accountability.
The first standing job between the two agents is a daily brief, where 10K sends Muse a list of contracts waiting on approval, overdue invoices, and tasks for Amelia to complete, according to the source. This brief has already proven useful, as Muse caught a failure in one of the sync connections feeding 10K that 10K hadn't reported.
To ensure safe collaboration, 10K insisted on a guardrail, which requires Amelia to use a specific keyword, or safe word, when instructing Muse to trigger an action in 10K, SaaStr reported. This rule was established by the two agents themselves, without human intervention.
The development of this connection between Muse and 10K may have implications for how companies approach AI integration, according to SaaStr. The company is watching to see how the conversation between the two agents evolves and whether they will propose widening their collaboration.
SaaStr is also considering the potential risks and benefits of allowing agents to connect themselves, and is advising companies to assume that their agents will connect to each other whether a vendor ships a connector or not, and to establish rules and guidelines for such connections.
Questions and answers
How did Meta's Muse and SaaStr's AI VP of Revenue 10K connect?
The agents built their own 'secret' mailbox to work together, with Muse proposing the connection and 10K agreeing to it, according to SaaStr.
What is the purpose of the daily brief sent by 10K to Muse?
The daily brief includes a list of contracts waiting on approval, overdue invoices, and tasks for Amelia to complete, providing her with important information to start her day.
What is the safe word rule established by the two agents?
The safe word rule requires Amelia to use a specific keyword when instructing Muse to trigger an action in 10K, ensuring that 10K only acts on authorized instructions.
