Tracking Cursor and Claude Code work in one issue tracker
A common state of the world in mid-2026: one developer drives Claude Code, another lives in Cursor, a third is experimenting with Codex, and a couple of background agents run on a schedule. Every one of them produces real work — branches, diffs, pull requests — and every one of them produces it in its own silo. The team’s issue tracker, meanwhile, knows nothing about any of it. Status updates happen in standup, by memory.
The instinct is to build a dashboard that aggregates the agents. That is the wrong shape. You do not need a fourth surface to watch the other three; you need the agents to report into the place the work already lives — the issue. This post is about how to make that happen without rewriting how each tool works.
The common protocol: MCP
The reason this is tractable is that Claude Code, Cursor, and Codex all speak the Model Context Protocol (MCP). Instead of writing a bespoke integration per tool, you expose your tracker as an MCP server and point each agent at it. The agent gains a small set of tools — fetch the issue it is assigned, read its context, post progress, attach output — and uses them the same way regardless of which client it runs in.
Hiveship ships an MCP server for exactly this. You register an agent in the workspace, generate a scoped bearer token, and drop the connection config into the agent’s MCP settings. The setup guide has copy-paste configuration for the common clients. From the agent’s perspective, Hiveship is just another set of tools it can call; from your perspective, every agent now writes back to the same board. The MCP server is free on every plan — see the pricing page for what each tier includes.
What “reporting in” actually looks like
Once an agent is connected, the issue becomes the system of record for its work. As the agent runs, Hiveship records an activity stream on the issue: the context it pulled, the changes it made, the diffs it produced, and the pull request it opened. Those events stream to the board in real time over SSE, so when you are watching an issue you see the agent move through its steps live — no polling three separate tools to assemble a picture.
Because the work lands on the issue, it also lands in everything that hangs off the issue: the board column, the project, comments, and — through the native GitHub integration — the linked pull request and its status. A reviewer opens one issue and sees the whole arc, whether the contributor was a person in Cursor or an autonomous agent on a cron.
Keeping it safe across multiple tools
Running several agents across several tools multiplies the number of credentials floating around, which is exactly when scoping matters. Each Hiveship agent token carries a capability tier — read, session, or workspace — and the API fails closed on any route that has not explicitly granted the agent’s scope. So you can give a Cursor-driven agent a session-scoped token that only lets it touch the one issue it is working, and reserve broader workspace access for a trusted orchestrator. If a token leaks or an agent misbehaves, the blast radius is bounded by the tier you handed out. The full model is in the agents documentation.
Why one tracker beats three dashboards
The deeper reason to consolidate is that a tracker is not just a viewer — it is where you plan, prioritize, and review. When agent work lives in separate tool dashboards, you can watch it but you cannot manage it: you cannot triage a Cursor task and a human task into the same sprint, or review an agent’s PR next to a colleague’s in the same queue. Pulling everything onto issues means the agents participate in your real process instead of running beside it.
And because Hiveship’s pricing is a flat per-workspace fee with no per-agent charge, connecting a fourth or fifth agent is free at the margin — the cost of consolidating does not scale with how many tools or agents you wire in. See the pricing page for the current plan limits.
Getting started
Pick the agent you use most — say, the one in Claude Code — register it, generate a session-scoped token, and assign it one issue. The Free plan includes one agent, so trying delegation and watching the activity stream fill in costs nothing; connecting through the packaged MCP server is free on every plan when you’re ready to wire up your clients. Once you trust the loop, add the next tool.
See the agent-first board, or jump straight to the MCP server setup.