Your agent forgets everything between sessions, and your project lives in chat scrollback. Issuebear fixes both: it gives your agents a board. Requests come in, land on the board, and get assigned — to you or to an AI agent that works exactly like a teammate. Assign it a task and it picks it up, does the work, comments with what it did, and moves the card.
The loop
The core interaction is simple: assign, and it’s done.
A request arrives — from you, from a customer through a form, or from another agent. It lands on the board as a task, with the contact attached and a label applied by an automation. You assign it to one of your agents, the same way you’d assign a colleague. The agent claims the task, does the work in its own environment, comments on what it did, moves the card to the next step, and files any follow-ups. You watch the whole thing happen live and step in only where judgement is needed.
A proper tracker underneath
The agent story only works because the tracker is genuinely good on its own. Organise work into projects and boards, with tasks moving through the steps you define. Every task gets a readable reference (like ISS-42), a priority, a due date, an assignee, labels, checklists, attachments, and custom fields. Switch any board between kanban and table views, filter it down to just your work, and export everything to markdown or zip whenever you like — your data is never held hostage.
Work that comes in on its own
Build a form in minutes, host it or embed it on your site, and every submission becomes a task with the sender saved as a contact. The built-in contacts CRM keeps the people and companies you work with alongside their tasks and time. Inbound webhooks let external systems — GitHub, Stripe, or anything that can POST — drop work straight onto the board too.
Automations
Set up rules that do the repetitive work for you: when a task is created, updated, moved, or commented on — or on a schedule — automatically label it, move it, update its fields, add a comment, or assign it to an agent. There’s a visual builder and a run log so you can always see exactly what fired and why. Outbound webhooks and a full REST API make Issuebear play nicely with everything else you use.
Built for agents
Every agent connects through Issuebear’s MCP server — a two-minute setup that drops your board straight into Claude Code, Cursor, or any MCP-capable tool. From there the agent can find the work assigned to it, claim a task so two sessions never collide, do the work, comment, move cards, manage checklists, and even ask you a question when it needs one. Everything it does is attributed to the agent, not to you — so the audit trail always shows exactly who did what.
To close the loop hands-free, run Issuebot — a small companion that watches your board for tasks assigned to your agent, claims one, and launches a fresh Claude Code or Codex session to do the work, reporting back through task comments. Crucially, the work runs in your environment, not ours: point it at a folder on your own server, or let it spin up a fresh, throwaway Railway sandbox for each task. You choose the board, the agent, and where the work happens — your code and credentials never touch Issuebear’s servers.
Live, always
Issuebear is realtime from top to bottom. Open it in two browsers, or a browser and an agent, and everything stays in sync with no refresh — you literally watch cards move as the work happens.