About KnownFix

Known breaks. Known fixes. Paste the error your AI agent harness just printed and KnownFix finds the matching known issue, with the workaround people quoted, the versions affected, its status, and a link to the source. It covers 14 harnesses: Claude Code, Codex CLI, Gemini CLI, GitHub Copilot CLI, Kiro, Amazon Q Developer CLI, Hermes Agent, OpenClaw, OpenCode, Goose, Cline, Aider, Continue and Omarchy.

Agents can use the same data without installing anything: a static JSON API and a context pack per harness. See For agents.

The fixes are scattered across thousands of GitHub issues and comments: updates break gateways, daemons stop restarting, auth tokens expire at 3am. KnownFix collects them in one place, and it's especially useful if you run agents always-on on a Mac Mini, mini PC or home server.

What's here

• Fix an error: paste an error and get the closest known issues with their quoted workaround. Matching runs in your browser over the public JSON. Your text is never uploaded.

• Known issues: open and recently closed GitHub issues from each harness's public repo, filtered to install and update breakage, crashes and hangs, gateway and daemon problems, auth, and headless setups. Ranked by how many people hit them. Refreshed nightly.

• Status: ALIVE / DEGRADED / DOWN for every harness (see how status works).
• Weekly email: a short "what broke this week" summary.

How it works

Everything is generated automatically. A nightly job pulls issues from the GitHub API, checks vendor status pages, and rebuilds this static site. Severity labels are automatic estimates from labels, keywords and engagement. They are not official ratings from the projects. Always check the linked issue.

How status works

Each harness panel shows ALIVE, DEGRADED or DOWN, and the worse of two signals wins:

1. Official status page, used where the vendor has one for the hosted service the harness depends on. We read its public JSON (Statuspage components.json, Google Cloud incidents.json, AWS Health currentevents, Better Stack index.json, Gatus endpoints/statuses). Mapping: operational → ALIVE; degraded performance, partial outage or maintenance → DEGRADED; major outage → DOWN.
2. Issue spike: GitHub issues whose title matches crash OR outage OR broken OR hang OR regression OR unusable opened in the last 48 hours, compared with the average per 48 hours over the previous 30 days. DEGRADED needs at least 4 such issues and 2.0× the baseline. DOWN needs at least 8 and 4.0×.

Thresholds live in config.json under status. Hover a pill to see the numbers behind it. If a status page can't be reached, the panel says "issue signal only" rather than guessing. The issue spike is a heuristic: it catches a bad release quickly but can also fire on a burst of unrelated reports.

The sparkline under each status shows tracked open issues per day for the last 30 days. Solid bars are daily snapshots taken at build time, stored in data/history.json. Lighter bars are backfilled from the created and closed dates of the issues we track. That backfill undercounts earlier days, because issues closed before the lookback window, or that never made the tracked set, aren't included.

Independence

KnownFix is an independent project. It is not affiliated with or endorsed by Nous Research, OpenClaw, Anthropic, OpenAI, or any person mentioned on the site. Product names belong to their owners.

Spotted something wrong? Corrections are welcome.

Looking for the nightly picks of agent news from X? They moved to the sister site, AgentRags.