Manual
for whoever installs, configures and gets pagedOne page per screen: what it decides, what has to be registered first, and what shows up when it works. Where the decision is meulog's and not yours, the entry says which one and why.
Getting started
- Sending errors to meulog the six doors in, from SDK to e-mail
- What meulog never stores sanitization on the way in: what goes by name and by shape
Investigating
- Errors the list that opens the day: one error per row, grouped
- A single error diagnosis: stack, curve, evidence, who took it
- Administering an error status, owner, merge, regroup, customer-facing text
- Occurrences every event, one per row, ungrouped
- A single occurrence one whole event: context, attachments, neighbours
- Customers who is suffering, and from what
- Code where in the code the place breaks most
- Searching by expression the filter you can paste in chat and save
- Dashboards the ones you build, and the ones that come ready
Records
- Projects who reports, with DSN, quota and platform
- Search clusters logs already indexed somewhere else
- Repositories where the AI reads the code around the error
- Source maps translating minified front-end stacks
- Flows the belt: where it comes in, what goes out, what is kept
- Parts and records every installed plugin and where each one fits
Watching
- Watched URLs the address, the pace, and what is expected of it
- Uptime the state of each URL right now, and the history
- Performance how long each operation takes, and since when
- Scheduled jobs the job that runs alone, and the alert for a missing signal
- Deploys the versions that shipped, and the verdict on each
- AI cost what each agent spent, and why
- Machine resources memory, CPU and disk of where meulog runs
Notifying
- Alert rules who gets told, about what, and when to go quiet
- Grouping rules when meulog merges or splits the wrong way
- What the AI suggests the proposal queue, and what each one does if approved
- Insights the period summary written by the analyst
- AI agents replacing the prompt of a task without touching the project
- Configuring in plain words describe what you want; the AI writes a blueprint and you approve it
- Prompt tests proving a prompt change did not break what worked
Operating
- Settings time zone, customer label, retention, purge, backup
- Processes the health of meulog, the queue and what is slow
- People who signs in, with which role
- Accounts one company per database, and how to switch
- Debug prove the belt moves, without waiting for a real error
- The read API read the errors from outside meulog, with the project token
- MCP: errors in your editor the four tools your editor agent uses
- Annotations what happened outside the software and explains the chart
- Evidence where screenshots and attachments end up