Let's do an honest agent roast. If you've tried using an AI agent in real work, tell us about a time it didn't work.
What were you trying to get done? What did the agent do instead? What did you have to redo or clean up? Stories from sales, operations, research, support, coding, or anywhere else are welcome.
Maybe it lost context between tools. Maybe it lacked permission, made something up with confidence, or broke in a completely different way. No polished demos, no pitches, and no need to have a fix. I'm curious which failures keep showing up when agents leave the demo and enter daily work.
What's your most memorable agent fail?
Thanks, @genedai Gene!
Yes, Shared Playbooks give your team's Skills and custom prompts a shared home.
A teammate can connect another supported agent to the same space, give it access to the relevant material, and ask it to consult that Playbook before starting a task. They can also copy saved prompts from the UI into their agent. For example, keeping a content review process and its prompts together lets a different agent read the same working instructions.
@seey Review changes before they go live is a good one. I’d want that if an agent is touching real business data.
@kate_ragotte19 Yes, Kate. With real business data, being able to see what an agent changed before accepting it makes a difference.
Thanks for the comment!
@seey @kate_ragotte19 Kate, that’s exactly where I’d turn on review. An agent can submit a proposed update while the current business record stays as it is. A person can inspect the changed fields, request a revision, or approve it. That makes the approval step part of the workflow, rather than discovering a bad update later in the history. What kind of business data would you put behind that gate first?
@genedai Yes. A Playbook can be a Skill with its instruction file and references, or a custom prompt saved under a node’s Agent prompts. Both appear in the workspace’s Playbooks view. Once another supported agent is connected to that workspace, it can find and read the relevant Playbook, subject to its access permissions.
For example, I’d save a repeatable code-review procedure as a Skill and a shorter “prepare this week’s launch post” instruction as a custom prompt. The team maintains each once instead of rewriting it for Codex, Claude Code, and the next agent. Here’s the Playbooks guide: https://busabase.com/docs/agent-playbooks
Which working method would your team reuse first?
@seey @mrkelly Curious how this compares with existing databases and agent memory solutions ?
Busabase focuses on multiple people and multiple agents maintaining the same shared data.
Think of it as: 【People : Agents : Data = M : N : 1】
Compared with a database, our focus extends beyond storage and queries to how people and agents maintain data together.
Compared with agent memory, our focus is on business records, docs, and working instructions that a team maintains as shared context for multiple agents. It doesn't automatically import or remember every agent conversation; it holds what you or your agent explicitly save.
@dipanshu_kushwaha5 Thanks for the comment!
@seey @dipanshu_kushwaha5
You can see a comparison in our open-source repo: https://github.com/busabase/busabase
Hey Product Hunt — Kelly here.
We kept seeing this with customers: a new project or a different agent often meant rebuilding context and instructions from scratch. Business users were juggling local files and agent setups, sometimes having to learn tools like Git just to keep things organized.
We believe agents should be stateless, with company knowledge and work kept in a shared base. Business users should be able to share that context across agents without becoming AI experts.
That's why we built Busabase — The Base for Agents — a shared workspace for data, docs, skills, and apps that people can see and manage.
Connect agents like Claude Code, Grok Bot, Muse, OpenClaw, Manus, Hermes, Buda or Codex via MCP, OpenAPI, or CLI. You decide what each agent can read or change, and which updates need review. You can also see what changed and who changed it.
With your agents and tools connected, you can automate workflows like:
• Software development — agents turn support feedback into tasks, write and test the code, deploy updates, and use new feedback to start the next cycle.
• SEO — agents monitor traffic and search trends, use the results to improve the content plan, write the next article in a shared CMS, and repeat.
• CRM — agents use product knowledge and shared business data to find relevant prospects, reach out, and follow up, keeping customer records, visit notes, and tasks up to date.
One customer keeps its company materials and project updates in Busabase. Now a manager can ask an agent for the latest status instead of going from teammate to teammate.
Try it with one project: add its updates and instructions, connect an agent, and ask, “What changed this week?”
Busabase is open source. You can self-host it, use Personal Desktop, or get started with Cloud.
If your team uses several agents, I'd love your feedback: what would Busabase need to handle for you to use it every day?
@mrkelly How do AI agents share the data with each other???
@vipul_kumar1280 Vipul, the agents don’t pass their private chat histories to each other. They connect to the same Busabase Space and read the project records, docs, and working instructions your team chooses to keep there. For example, Codex can propose a task update; once it’s approved, Claude Code can read that updated record on its next task. Each agent still needs to fetch the context and write back useful results. What would you want your agents to hand off first?
@vipul_kumar1280 Just connect them into one shared Busabase. Maybe it is not so called "share the data with each other", they share the same data
Our team uses both Codex and Claude Code. What's the simplest way to give them the same project brief and coding conventions without copying everything into each chat?
@matthewwei Matthew, I'd start with one project: save the brief as a Busabase Doc and the coding conventions in a shared Playbook. Connect both Codex and Claude Code to that workspace with read access.
Then add a short instruction to each tool's project instructions: “Before starting work, read this project's brief and coding conventions in Busabase.” That gives you one copy of the full content to maintain.
Where do those materials live today, in your repo or somewhere like Notion? I can help sketch a small setup around what you already use.
@matthewwei Personally, I think AGENTS.md for Codex and CLAUDE.md for Claude Code are great ways to avoid repeating the same instructions or explaining business context every time you start a new session.
You can also include instructions that tell the agents when to pull relevant business context from Busabase.
For example:
This way, both agents know when to retrieve business context, where to find it, and how to use it — without you having to repeat yourself in every session.
Here's what that looks like in the AGENTS.md file of one of my projects 👇
This has been especially useful for our marketing team. Once teammates clone the project locally, they can all reference the same product brief stored in Busabase. So even when we're running different AI tasks across different sessions or agents, we're all working from the same single source of truth, keeping our outputs consistent and grounded in the same facts.
@matthewwei Matthew, Seey gave you the practical setup. The part I’m genuinely excited about is what comes next: install the Busabase Skill in both Codex and Claude Code, connect them to the same Space, and give them one shared home for the brief, coding conventions, and tasks. Either agent can read that context and leave work the other can pick up.
We use this ourselves for much more than project plans: our CMS articles, SEO keywords, engineering changelog, and product docs live there too. The agent can be new; the company’s context doesn’t reset. That’s the Superman moment for me. It feels like a fresh agent can step in with the team’s accumulated working knowledge. I’m genuinely excited every time I see that handoff work.
Agent setup: https://busabase.com/docs/agents-getting-started
Playbooks: https://busabase.com/docs/agent-playbooks
What would you want both agents to know on day one?
Looks cool, I want to ask if an agent should read project records but only propose CRM changes, how would you set that up? What can the reviewer see before accepting a proposal?
@limxn6 Bertram, I’d keep the project material and CRM in separate nodes. Give the connected identity read access to the project material and proposal access to the CRM Base, then cap the agent’s connection at changeRequest. It can read what it needs and submit a CRM change, but it cannot merge that proposal itself. In the Inbox, the reviewer can inspect the proposed fields against the current record, see who submitted the change, and approve or request changes. The connection guide is here: https://busabase.com/docs/mcp
Amazing idea!
Have to deal with this a lot now. Setting up each agent is a challenge. Would give this a shot today!
Hey Product Hunt! 👋 I’m another maker on the Busabase team.
Kelly shared why we built Busabase. I’d like to highlight a few features that help teams put it into practice.
For us, sharing information is only part of the job. Teams also need to preserve how they work, understand what people and agents contributed, and turn accumulated knowledge into something useful day to day.
Here are four features I’m especially excited about:
Shared Playbooks: Give your team’s Skills, procedures, coding conventions, and custom prompts a home in one browsable catalog. Connected agents can consult the relevant instructions, and teammates can copy saved prompts for their next task. A useful method developed by one person becomes something the team can reuse.
Team Insights: See contributions recorded in Busabase Cloud, including changes, proposals, reviews, and activity trends. Owners and admins can view member contributions and their human and agent sources, helping them understand how shared work is being maintained.
Instant AirApps: Ask an agent to turn workspace data into a dashboard, briefing, or internal business app. It runs directly inside Busabase and stays connected to the underlying records.
Agent-ready Templates: Start with Bases, Docs, AirApps, sample data, prompts, and instructions packaged together, then adapt that setup to your team’s workflow.
Which working method would you put into a shared Playbook first—a coding convention, a content process, or something else? I’d love to hear your examples.