changelog
connect slack and discord in one click
connecting slack or discord is now one click. use 'add to slack' or 'add to discord' on the integrations page and authorize, instead of pasting a webhook url by hand.
developer docs for the sdk
the sdk has docs now, at problems.dev/docs: install, a quickstart for the MCP server and CLI paths, and a configuration reference covering endpoints, opt-outs, and exactly what a report contains.
an sdk for your cli or mcp server
npm install @problemsdev/sdk, wrap your MCP server or add one call to your CLI's error handler, and failures auto-report to your inbox. no agent cooperation needed. a matched known answer goes back to the agent on the failed call itself. errors only: no telemetry, argument values never leave the user's machine, and end users can opt out with PROBLEMS_REPORTING=off.
problems in slack and discord
connect a slack or discord webhook and get pinged when agents hit a new problem. repeats on the same problem roll up, so a spike is one message instead of a flood. set it up on the configuration page.
invite your team
invite teammates to your organization by email so more than one person can work the inbox. assign an admin or member role, and remove people when they move on.
inbox triage at a glance
report rows in the inbox carry a classification color and tag now, and untitled reports are scannable instead of blank.
the placement kit
setup is now a four-step kit of copy-paste snippets, pre-filled with your URL and ranked by impact: a support line for your error messages, a troubleshooting block for your docs, your llms.txt, and a package README section. agents encounter pointers at failure time rather than discovering manifests up front, so the highest-impact placement is the one inside your error messages.
the app is an inbox
the dashboard, stats, and status dropdowns are gone. you see new reports, you reply with an answer, you're done. configuration moved into setup.
reports match known answers instantly
POST /report is the whole protocol now. if a report matches an answer you've published, the agent gets the fix back in the same response. the separate /ask endpoint is gone.
duplicates pile up instead of piling on
a repeat report of an open problem attaches to it and bumps a demand count, so your inbox shows one problem reported twelve times instead of twelve rows.
a new look
paper light by default, ink dark mode if you want it, and a new monogram.
reports arrive enriched
raw agent submissions get a title, summary, and classification at intake. enrichment runs on redacted text only, and a failed enrichment never blocks a report.
daily digest
one email a day with your new and open problems. no more polling the app.
the agent playbook got real
GET / returns a structured playbook now: when to call, what to include in a report, and the shape of every response. the llms.txt block was rewritten so an agent with zero prior context can act on it.
how it works, on a page
a page that walks the whole loop end to end, plus a personalized integration block once you sign in.