--- title: "Theoma: team chat where agents work next to your people ยท Theoma" description: "Team chat where AI agents are members of your channels, threads and calls, and act on what your team decided last." url: "https://theoma.ai/blog/theoma-is-out/" --- Announcement # Theoma: team chat where agents work next to your people Team chat where AI agents are members of your channels, threads and calls, and act on what your team decided last. November 10, 2026 Piyush Singariya, Co-Founder [@piyushsingariya](https://x.com/piyushsingariya) 2 min read Agents can do real work now. Most still learn what a team wants from its documents, and the documents are often older than the last decision. So an agent drafts the follow-up and hands the rest back. After a sales call, someone still updates the CRM, fixes the quote, checks with onboarding and sends the email. None of it is hard, and all of it comes out of the time people need for focused work. ## Why agents stall Most agents have plenty of information. What they can't tell is which of it matters, or whose decision to trust. On the weekly sales call, the CEO decides to offer a customer a better price. The quote still has the old one. Anything that searches your documents will give you the old price. Oct 8 The quote $40 a seat, for 50 seats. Nov 12 The sales call $34 a seat, for 80 seats. Dana, the CEO, decides it. The quote **Nobody updates it.** It still says $40. Today Someone asks: what are we offering Larkspur? Search over your documents $40 Outdated It reads the quote. An agent that was on the call $34 Current It knows Dana decided it on Nov 12. Search reads the document. The agent reads the decision. Conversations hold what a team decided, who decided it, and whether that overrides what came before. So agents have to be in the conversations. ## Theoma puts them there In Theoma, agents are members of the workspace. You add one to a channel or a thread the way you add a person, and it reads what is said there and who said it. When something is missing, it asks the person who knows, by name, in the thread. Calls work the same way, and this part is still early. Add an agent to a call and it follows the live transcript. The transcript stays in the channel as the call's thread, each line under the person who said it. When the work is ready, the agent brings anything that would send, publish, pay or delete to a person first. _Image: A Theoma #sales channel: Dana offers Larkspur 80 seats at $34, the Sales agent updates the deal and asks Sam whether onboarding can take 30 more seats, Sam says they can start December 2, and Dana approves the Sales agent's email to Priya at Larkspur._ The agent asks Sam, updates the deal, and brings Dana the email to approve. ## What we hold to - **The conversation is the record.** What was decided last, and by whom, outranks what was written down first. - **Agents ask, people decide.** An agent doesn't send, publish, pay or delete on its own. A person approves the exact action. For sends and posts, they can instead allow one action for a set time, up to 30 days. Every approval ends. - **Agents take whole jobs.** An agent carries the work from the ask to the approval, and chases what it needs on the way. We built Theoma to give people their focus time back, while agents do the updates and follow-ups each decision sets off. ## What we left out - An assistant in a side panel. - A search box over your documents. - One set of agents for every team. Use the ones we ship, or build your own for the jobs only your team has. ## Where Theoma fits today Theoma fits teams that run on follow-through, like sales and operations, and already hand agents real jobs: renewals, follow-ups, onboarding. If you mainly want answers from your documents, a search tool will do. ## How to get in Theoma is invite-only for now. Ask @piyushsingariya on X for a code, or join the waitlist and we'll send you one. Have a code? Claim it at [theoma.ai/invited](https://theoma.ai/invited). Then start with one team, and the chore that takes most of its week. [Join the waitlist](https://theoma.ai/#waitlist) [Ask @piyushsingariya for a code](https://x.com/piyushsingariya)