September: getting our agents in a row
I spent September using Annsa to build Annsa. My agents across Claude Code, Codex and Cursor worked from one list, and here are the 10 things we added along the way.
It's spring in Sydney. The weather has been glorious, the whales are migrating and I'm back on iced coffee in the sunshine.
For the past month, I've been using Annsa to build Annsa.
Annsa began as a tool for bringing together customer feedback. Dogfooding it, treating myself like Customer Zero, showed me it had to do more than that.
Using your own product every day changes what you notice.
In July and August, I was struggling to manage the full picture: customer feedback, my own ideas, the roadmap and everything my agents found while they built. Anyone who has worked with me knows I have a hard time getting to a top 3. There's always another problem worth solving.
Meanwhile, my husband, my mother and both my brothers have been giving me a hard time for not getting out to market enough. Ha! I used to do that. But my agents have taken me hostage.
So September became about getting us all working from the same direction.
My agents across Claude Code, Codex and Cursor worked through 370 priorities across 5 releases. So much for a top 3. I see you, 370 things.
Special mention to Grok, which didn't build a thing but critiqued absolutely everything.
Every day, I'd ask the agents what worked, what didn't and whether they'd recommend Annsa. Then we'd work through the feedback to make the experience better, for them and for me.
Their feedback was one perspective. Using Annsa myself showed me where I was still repeating decisions, explaining context again or trying to work out what needed my attention.
Agents often asked for guidance or decisions, so we baked that into the product. Quite frankly, so I could stop repeating myself, let the agents work overnight and get a walk in each day.
That's the bit I wanted. Make a call, keep the reason and let the next agent build from it.
So if you're like me, building a lot, trying to choose what matters and wanting your agents to work from the same direction, here's what's new in Annsa.
10 things we added in September
1. Every signal in one list
Your roadmap, ideas, agent findings and customer feedback come together, with their sources kept distinct.
Read the 18 Sep release note →
2. Rank the work against your bets
Name the outcomes you're betting on, and see priorities grouped and ranked under the bet they serve.
3. Your agents can work through Annsa
Pick up work, write their part of a spec, hand over and record what shipped, all through MCP.
Read the 14 Sep release note →
4. Answer your agent from your terminal
When a call is yours, the agent asks in Needs you. Your answer stays with the priority and the spec.
Read the 17 Sep release note →
5. Answer once. The next agent can read it
Keep decisions with the work, and save rules that should guide future work too.
Read the 18 Sep release note →
6. See what was ruled out, and why
Whoever picks up a priority next can see the approaches already considered and the reasons they were set aside.
Read the 14 Sep release note →
7. Different words can point to the same problem
Related feedback comes together on one priority, with the original customer words still attached.
Read the 17 Sep release note →
8. Tell the people who asked
Share back brings together the customers behind a priority. Review and send the follow-up; their replies land on the work.
Read the 17 Sep release note →
9. Shipped doesn't mean forgotten
When new customer feedback reopens shipped work, it returns to Needs you with the previous outcome on its row.
Read the 29 Sep release note →
10. Turn the work into a release
Bring a period of shipped work together as one release. It's how I put this month's changelog together.
Read the 17 Sep release note →
That's September. If you want your agents working from the same direction, connect Annsa to the tool you already build in.
Give your agents shared direction.
claude mcp add --transport http annsa https://app.annsa.ai/mcp