One thing I forgot to mention that blew our own minds:
We built CodeBake using CodeBake.
From the very first ticket → final deploy, our AI agents were real participants on the board (pulling tasks, updating statuses, writing summaries). Zero manual PM busywork for the entire project.
That’s when we knew this wasn’t just a side project — it actually removed an entire category of work for us.
I'm Ben from https://MisfitLabs.vc — small team that’s been building healthcare software together for the last 12+ years.
We built CodeBake because we were frustrated with the "flexibility trap" of traditional PM tools. After 30 years of using Jira, Asana, and Linear, I realized we were spending more time managing the tool than doing the work.
We've been using CodeBake internally for most of 2025 (eating our own dogfood), and we realized that project management should be a byproduct of work, not a separate activity.
THE CORE IDEA: A BUILT-IN MCP SERVER
If you use AI agents like AugmentCode (our favorite), Claude Code, or Cursor, the AI already knows what you are building. It knows the context. So why are you manually tabbing over to a browser to update a status or write a summary?
CodeBake comes with a Model Context Protocol (MCP) server that lets your AI agent become a first-class participant on the board. You can stay in your editor and just say things like:
"What's my next task?"
"Let's work on CB-233."
"Update the CodeBake task with a description of the solution we're implementing."
"Mark this task as completed."
Your agent handles the API calls, context fetching, and status updates automatically.
CLIENT MANAGEMENT & AI REPORTS
Another massive pain point was keeping clients in the loop without letting them mess up the board.
We built a dedicated "Presentation View" specifically for reviewing tasks during online meetings, along with unlimited read-only "Guest" roles so clients can see progress without edit access.
Since the system knows exactly what happened, clients have access to one-click AI summaries of what was accomplished in any specific time range. It ends the Friday 4 pm death march of writing status reports.
THE TECH STACK:
We are 100% Elixir and Phoenix LiveView on the backend, backed by Postgres.
Why LiveView? We needed "optimistic UI" speeds to keep the board in sync with high-frequency AI updates. LiveView pushes state over the socket instantly, meaning when your agent updates a task, you see it change in real-time without a page refresh.
BETA:
Opening the first 400 external spots to HN right now — use code HNBETA2025 to skip the waitlist (then it closes again).
Happy to talk MCP details, LiveView performance, pricing concerns, whatever. I’ll be here all day.
One thing I forgot to mention that blew our own minds:
We built CodeBake using CodeBake.
From the very first ticket → final deploy, our AI agents were real participants on the board (pulling tasks, updating statuses, writing summaries). Zero manual PM busywork for the entire project.
That’s when we knew this wasn’t just a side project — it actually removed an entire category of work for us.
Hey HN,
I'm Ben from https://MisfitLabs.vc — small team that’s been building healthcare software together for the last 12+ years.
We built CodeBake because we were frustrated with the "flexibility trap" of traditional PM tools. After 30 years of using Jira, Asana, and Linear, I realized we were spending more time managing the tool than doing the work.
We've been using CodeBake internally for most of 2025 (eating our own dogfood), and we realized that project management should be a byproduct of work, not a separate activity.
THE CORE IDEA: A BUILT-IN MCP SERVER
If you use AI agents like AugmentCode (our favorite), Claude Code, or Cursor, the AI already knows what you are building. It knows the context. So why are you manually tabbing over to a browser to update a status or write a summary?
CodeBake comes with a Model Context Protocol (MCP) server that lets your AI agent become a first-class participant on the board. You can stay in your editor and just say things like:
"What's my next task?"
"Let's work on CB-233."
"Update the CodeBake task with a description of the solution we're implementing."
"Mark this task as completed."
Your agent handles the API calls, context fetching, and status updates automatically.
CLIENT MANAGEMENT & AI REPORTS
Another massive pain point was keeping clients in the loop without letting them mess up the board.
We built a dedicated "Presentation View" specifically for reviewing tasks during online meetings, along with unlimited read-only "Guest" roles so clients can see progress without edit access.
Since the system knows exactly what happened, clients have access to one-click AI summaries of what was accomplished in any specific time range. It ends the Friday 4 pm death march of writing status reports.
THE TECH STACK:
We are 100% Elixir and Phoenix LiveView on the backend, backed by Postgres.
Why LiveView? We needed "optimistic UI" speeds to keep the board in sync with high-frequency AI updates. LiveView pushes state over the socket instantly, meaning when your agent updates a task, you see it change in real-time without a page refresh.
BETA:
Opening the first 400 external spots to HN right now — use code HNBETA2025 to skip the waitlist (then it closes again).
Happy to talk MCP details, LiveView performance, pricing concerns, whatever. I’ll be here all day.
Thanks! - Ben