Last year we built agents coordinating agents. Users asked for it. Demos looked great. Then we watched them not use it.
Nobody knows what an agent actually is, and nobody wants to configure them. What they wanted was simpler: "pull reports from my CRM and send them to me every day at 8am."
We scrapped it and built V3 around a ginle interface: describe what you need, it gets built. The agent (Nira) helps you build - you don't build agents.
Under the hood we're orchestrating across MCP servers - intent classification, dynamic tool loading, state management, scheduling. MCP registry is open source: https://github.com/NimbleBrainInc/mcp-registry
Free tier has all integrations unlocked.
Happy to answer questions about the architecture or why we killed agents.
jackfranklyn•1h ago
Started with "configure your coding rules" - let users set up how transactions should be categorised. Made sense architecturally. Nobody used it. They'd stare at the config screen and give up.
What actually worked: learn from their existing data. Pull their historical GL, reverse-engineer how they already code things, then just do it that way. No configuration, no "agent" concept. Upload bank statement, get coded transactions back.
The insight that killed our complexity was realising bookkeepers don't think in rules. They think in outcomes. "This should be coded like the last time we saw a payment from Stripe" not "if merchant contains 'stripe' and amount > 100 then category = payment processing fees".
Curious about your MCP orchestration layer - how do you handle cases where the described workflow is ambiguous? In accounting there's often multiple valid ways to code something.