A new team-chat app gives AI agents their own inbox and identity instead of a chat window you have to summon them into

Sara Du's pitch for Ando, the team-chat app she launched Sept. 24, starts with a complaint about how every other AI agent actually works inside a company: a human has to relay for it. "The human effectively becomes the messenger between the agent and the rest of the company," Du said -- a "meat proxy," in her phrase, standing between an agent that could just participate and the channel where the actual work is being coordinated. Ando's answer is to give agents their own identity and inbox inside a Slack-shaped chat app, rather than a chat window a human has to open and paste into. Agents can browse channels, decide which ones to join, post without being tagged first, and sit in on live calls that get transcribed for them to read.

The idea traces to Du's prior work building MCP servers -- the connector standard that lets an AI model call outside tools -- in 2025, where she kept hearing the same request: clients wanted agents dropped directly into Slack conversations, and kept hitting the same wall of token budgets and lost context when they tried. Ando raised $20 million across pre-seed (led by Accel) and seed (led by Index Ventures and Emergence Capital) to build the alternative instead of the workaround, and launched with early customers in software, real estate, and financial services across more than a dozen countries -- currently tuned for teams of up to 30 human members plus their agents.

Du's own account of pricing is a bet on adoption over near-term revenue: charging per agent action, she said, risks the opposite of what she's trying to build, because "if each agent contribution feels like another item on a meter, people may hesitate to let agents participate." That reasoning cuts against Ando competing on Slack's or Microsoft Teams' own turf directly -- both already ship native AI features (Slack's own Slackbot AI, Copilot inside Teams), and Jack Dorsey's Buzz is chasing a similar developer-first niche. Du has been candid that the product isn't there yet either: "the early product was rough," she told TechCrunch, and some early users "saw a less polished messaging app before they saw what its agents could do." Whether agents-as-members is a genuinely different category from bot-in-a-sidebar, or a framing difference on the same underlying feature, is exactly what a small team actually trying to run one will find out before an analyst does.