How Tomorrow Today wired GitHub and Marker.io into one Claude Code turn

The problem they were solving

Tomorrow Today's engineering team runs out of GitHub Issues: it's where work gets planned, prioritized, and closed. Bug reports submitted through the Marker.io widget already synced into GitHub, but the parts that actually help an engineer reproduce a bug — screenshots, console logs, network traces, DOM state — stayed inside Marker.io. Every triage meant tabbing between Claude Code and the Marker.io web app to assemble enough context to even start the fix.

‍

What they built

Tomorrow Today connected the Marker.io MCP server to Claude Code so the full bug context arrives at the moment of triage, not after a round trip to a second dashboard.

• Claude Code pulls the GitHub issue (conversation, ownership) and the Marker.io record (screenshot, console log, network trace) in a single pass and writes the fix in the same turn.
• Two-way sync via MCP: resolving an issue in GitHub flips the matching Marker.io issue to resolved, and vice versa. No more "closed in one place, still open in the other."
• Workflow runs from a markdown list of issues. Scott's actual prompts:

‍

"pull the new issues in this repo and add them to the issue list. if more context is needed to triage an issue, get more information from marker using the mcp"

‍

"mark completed issues rtt in github and resolved in marker. make sure to document the root issue and fix for each."

‍

Where they are now

Daily workflow. Claude Code is the single surface — GitHub for tracking, Marker.io for context, both reachable through the MCP without leaving the editor. Scott's team marks completed issues RTT (ready to test) in GitHub and Resolved in Marker.io as the closing step, with Claude documenting the root issue and fix for each one.

‍

Why this is worth stealing

• The bug context belongs in the same turn as the fix. Marker.io captures the screenshot, console log, and network trace. Once that's behind an MCP server, Claude Code can use it as input to its reasoning step instead of waiting for an engineer to fetch it.
• Two-way sync removes the close-state drift. A bug closed in GitHub but still open in Marker.io (or vice versa) is the kind of mismatch that nobody notices until a user re-reports the same bug a week later. MCP keeps both sides honest automatically.
• One markdown list, one prompt, one workflow. Scott runs the entire triage-to-RTT loop from a markdown checklist with two reusable prompts. No new dashboard or new tooling, only a thin AI layer over the systems his team already lives in.