Railway, Vercel, Netlify, Render and your own servers, searchable in one place.
Free plan · no card needed
Sources
Likely cause: the Postgres pool is full (max 10), so requests wait 30 s and time out. Raise the pool size or close the idle connections the cron job leaves open.
How it works
Connect where your apps run, search them together, and let AI read the error for you.
About two minutes
Paste a Railway token, add a Vercel or Netlify drain, point a Render log stream, or send JSON from any app.
Answers in under a second
One search across every app: words, levels, services, status codes, paths and request ids.
search
status:5xx path:/api/* -healthcheck
→ 38 lines in 3 services
ERROR POST /api/checkout 502 1840ms
ERROR GET /api/me 500 22ms
One AI credit
Pick the lines around an error and get the likely cause and the fix, in plain words.
Explain with AI
Cause: STRIPE_SECRET_KEY is a test key in production, so every checkout is refused.
Fix: set the live key on the service and redeploy.
Features
No more five dashboards and `railway logs` in a loop. Logyak keeps the lines, finds the ones that matter and tells you what they mean.
Follow a request from the edge on Vercel to your API on Railway and the worker on Render, by its id, in one list.
| 09:14:02.118 | web | INFO | POST /checkout 200 (edge, syd1) |
| 09:14:02.164 | api | INFO | creating order for cart c_41 |
| 09:14:03.902 | api | WARN | stripe: retrying after 429 |
| 09:14:05.377 | worker | ERROR | webhook signature mismatch |
Watch lines arrive as they happen, filtered to what you are chasing. Leave it open during a deploy.
level:>=warn service:api status:5xx path:/api/* "timeout" -healthcheck. Save the ones you use as one-click chips.
Railway, Vercel and Netlify edge requests with method, path, status and duration, next to your app's own lines.
The likely cause and the fix for the lines you pick. IPs, emails and secrets are masked before the model sees them.
Open a line and see the 20 lines before and after it from the same service, the way you would scroll a terminal.
A histogram by level over any range shows when errors started, before you read a single line.
Send logs with one POST from anything that speaks HTTP, and search them from your own scripts with an API key.
Invite teammates, keep clients apart in workspaces, and connect up to 100 sources on one plan.
From 3 days on the free plan to 30 days on Team. Old lines are removed every night, never before.
No agent, no sidecar. Each platform connects the way it was built to.
AI
Pick the lines around a failure. The AI reads only those lines, masked, and answers with the likely cause and the steps to fix it. Each answer costs a credit, and every plan includes some.
Picked lines
The API lost its database at 03:12 and every request after that failed. The lines show the host name resolving but the port refusing, which fits a database restart or a full disk rather than a network change. Check the database service's own logs at 03:12, then its volume usage; once it is back, redeploy the API so the pool reconnects.
IP addresses, emails and secrets are masked before the AI sees a line.
Security
Logs hold your customers' details. Here is exactly how Logyak handles them.
Your line
POST /login 401 user=ana@shop.com ip=203.0.113.7 token=sk_live_4f9...
Sent to the model
POST /login 401 user=[email] ip=[ip] token=[secret]
Masking runs on the server before any AI call. The masked text is also what the AI usage log keeps.
FAQ
A hosted place for your apps' logs. It collects them from Railway, Vercel, Netlify, Render and anything that can send an HTTP request, keeps them for your plan's retention, and lets you search, tail and explain them.
Both. The free plan takes 500 MB of logs a month from 3 sources and keeps them for 3 days, forever. Every paid plan starts with a 14-day trial and needs no card.
You paste a Railway token once. Logyak reads your app and HTTP logs through Railway's own API every 30 seconds, for the project and environment you pick. The token is stored encrypted and can be read-only.
The size of each stored line: its message and attributes. Lines dropped because you were over the allowance are not counted, and the dashboard shows how many there were.
New lines stop being stored until the month ends or you upgrade. You get an email at 80% and at 100%. Nothing already stored is lost.
Not yet. Logyak is for finding and understanding what happened. Alerts and on-call are planned as a separate product that will read the same logs.
Only the lines you pick, after IP addresses, email addresses, tokens and key=value secrets are replaced with placeholders on our server. Nothing is sent to AI unless you click Explain.
Yes, from the billing page, in two clicks. Your plan runs until the end of the period you paid for, then the free plan's limits apply.
Connect your first platform in two minutes. Free plan, no card.