This product was not featured by Product Hunt yet. It will not be visible on their landing page and won't be ranked (cannot win product of the day regardless of upvotes).
Local-first DevTools for Claude Code and OpenCode. Pinpoint prompt cache breaks, retry loops, and context bloat in under 60 seconds. Copy remediation rules to stop compounding token waste. Runs 100% locally. Code stays on your machine.
I am Krzysztof, a software architect and engineer. I run autonomous coding agents like Claude Code and OpenCode for daily engineering tasks.
During a multi-turn refactoring task, an agent session burned through prompt tokens after an early edit invalidated prompt cache prefixes for subsequent turns. In the same session, three consecutive tool failures repeated with identical arguments and consumed budget without making progress.
That was a single run. You don't build a profiler after one bad session. Across dozens of daily agent sessions, though, those hidden inefficiencies compound: invalid cache prefixes multiply prompt token spend, tool retry loops waste developer time, and bloated contexts degrade output quality.
Provider dashboards report total dollars without showing the prompt turns or tool calls where the session degraded. Enterprise APMs require cloud accounts, docker collectors, and sending proprietary code over the network.
So we built RunRay to diagnose local agent sessions the way DevTools diagnose browser applications.
How it works:
Auto-discovery: Run npx runray view. It scans local Claude Code (~/.claude/projects) and OpenCode logs on your workstation and opens the profiler in your browser.
Spend Spine and Leak Rail: A cumulative cost sparkline and context curve show the exact turn where spend spiked or caching broke. Clicking any point scrolls to that span in the Delegation Waterfall.
Deterministic findings: 12 diagnostic rules identify retry loops, cache prefix breaks, and context bloat, pairing each finding with a copyable remediation command or CLAUDE.md guard.
Local privacy: Runs on localhost without sending data over the network. Prompts and code remain on your machine. You can also export an Autopsy Report with runray export --anonymize to share in pull requests.
You can inspect your own sessions in under 60 seconds:
npx runray view
If you do not have local logs handy, run the scrubbed sample session:
npx runray demo
What it does not do yet:
RunRay currently parses Claude Code, OpenCode, and OpenTelemetry JSON logs. It does not yet support Codex, pi or Antigravity CLI logs, and it operates strictly as an on-demand local session profiler. Future runtime adapters and integrations will be prioritized directly from developer feedback.
Code, documentation, and discussions are on GitHub:
If you run npx runray view on a recent session, paste your 3-line Markdown Session Summary or link an anonymized report.html (runray export --anonymize) in the comments below. I will review your trace and reply with the specific CLAUDE.md or AGENTS.md remediation rule for your main token spike.
I would also value your feedback on two questions:
Which agent runtime would you like to see supported next?
What failure patterns or token spikes do you run into most often?
No comment highlights available yet. Please check back later!
About RunRay on Product Hunt
“Local-first DevTools for autonomous coding agents”
RunRay was submitted on Product Hunt and earned 0 upvotes and 1 comments, placing #443 on the daily leaderboard. Local-first DevTools for Claude Code and OpenCode. Pinpoint prompt cache breaks, retry loops, and context bloat in under 60 seconds. Copy remediation rules to stop compounding token waste. Runs 100% locally. Code stays on your machine.
RunRay was featured in Developer Tools (523.2k followers), Artificial Intelligence (483.9k followers), GitHub (41.5k followers) and Tech (636.9k followers) on Product Hunt. Together, these topics include over 441.2k products, making this a competitive space to launch in.
Who hunted RunRay?
RunRay was hunted by Krzysztof Lempka. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how RunRay stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Hi Product Hunt,
I am Krzysztof, a software architect and engineer. I run autonomous coding agents like Claude Code and OpenCode for daily engineering tasks.
During a multi-turn refactoring task, an agent session burned through prompt tokens after an early edit invalidated prompt cache prefixes for subsequent turns. In the same session, three consecutive tool failures repeated with identical arguments and consumed budget without making progress.
That was a single run. You don't build a profiler after one bad session. Across dozens of daily agent sessions, though, those hidden inefficiencies compound: invalid cache prefixes multiply prompt token spend, tool retry loops waste developer time, and bloated contexts degrade output quality.
Provider dashboards report total dollars without showing the prompt turns or tool calls where the session degraded. Enterprise APMs require cloud accounts, docker collectors, and sending proprietary code over the network.
So we built RunRay to diagnose local agent sessions the way DevTools diagnose browser applications.
How it works:
Auto-discovery: Run npx runray view. It scans local Claude Code (~/.claude/projects) and OpenCode logs on your workstation and opens the profiler in your browser.
Spend Spine and Leak Rail: A cumulative cost sparkline and context curve show the exact turn where spend spiked or caching broke. Clicking any point scrolls to that span in the Delegation Waterfall.
Deterministic findings: 12 diagnostic rules identify retry loops, cache prefix breaks, and context bloat, pairing each finding with a copyable remediation command or CLAUDE.md guard.
Local privacy: Runs on localhost without sending data over the network. Prompts and code remain on your machine. You can also export an Autopsy Report with runray export --anonymize to share in pull requests.
You can inspect your own sessions in under 60 seconds:
If you do not have local logs handy, run the scrubbed sample session:
What it does not do yet:
RunRay currently parses Claude Code, OpenCode, and OpenTelemetry JSON logs. It does not yet support Codex, pi or Antigravity CLI logs, and it operates strictly as an on-demand local session profiler. Future runtime adapters and integrations will be prioritized directly from developer feedback.
Code, documentation, and discussions are on GitHub:
Repository: https://github.com/RunRay/runray
Community & discussions: https://github.com/RunRay/runray/discussions
If you run npx runray view on a recent session, paste your 3-line Markdown Session Summary or link an anonymized report.html (runray export --anonymize) in the comments below. I will review your trace and reply with the specific CLAUDE.md or AGENTS.md remediation rule for your main token spike.
I would also value your feedback on two questions:
Which agent runtime would you like to see supported next?
What failure patterns or token spikes do you run into most often?