The only tool you need to know what is happening and how to fix it.
Go alertingexception-trackerissue-trackinglogsmetricsmonitoringobservability-platformopen-telemetryotelsession-replaytelemetrytraces
CLI command to MCP tool map
The knowledge resources show traceway CLI commands in their examples. Every command maps 1:1 to an MCP tool with the same parameters (flags become snake_case params). Use this table to translate.
| CLI command | MCP tool |
|---|---|
traceway projects list |
list_projects |
traceway exceptions list |
list_exceptions |
traceway exceptions show <hash> |
get_exception |
traceway exceptions occurrence <id> --recorded-at <t> |
get_exception_occurrence |
traceway logs query |
query_logs |
traceway endpoints list |
list_endpoints |
traceway endpoints requests <endpoint> |
list_endpoint_requests |
traceway endpoints show <id> --recorded-at <t> |
get_endpoint_request |
traceway endpoints chart |
endpoints_chart |
traceway endpoints slow <endpoint> |
get_slow_endpoint_config |
traceway tasks show <id> --recorded-at <t> |
get_task |
traceway ai-traces show <id> --recorded-at <t> |
get_ai_trace |
traceway sessions show <id> --started-at <t> |
get_session |
traceway traces show <id> --recorded-at <t> |
get_trace |
traceway metrics query --name <n> |
query_metrics |
traceway pages list / pages show <id> |
list_pages / get_page |
traceway alerts list |
list_notification_rules |
traceway monitors list / monitors show <id> |
list_checks / get_check |
traceway post-mortems list / post-mortems show <id> |
list_post_mortems / get_post_mortem |
Write server only (traceway-write mcp, or /mcp/write on the instance):
| CLI command | MCP tool |
|---|---|
traceway-write exceptions archive <hash> --yes / unarchive |
archive_exceptions / unarchive_exceptions |
traceway-write pages ack <id> / pages resolve <id> |
acknowledge_page / resolve_page |
traceway-write alerts enable <id> / alerts disable <id> |
set_notification_rule_enabled |
traceway-write alerts snooze <id> --minutes N |
snooze_notification_rule |
traceway-write monitors run <id> |
run_check |
traceway-write post-mortems create / update <id> |
create_post_mortem / update_post_mortem |
traceway-write endpoints mark-slow <endpoint> |
set_slow_endpoint |
Flag translation: --since/--from/--to are since/from/to; --page/--page-size are page/page_size; --recorded-at is recorded_at; --started-at is started_at; --min-severity is min_severity; --search-type is search_type; --order-by is order_by; --interval-minutes is interval_minutes; --metric-type is metric_type.
Differences from the CLI:
- Project selection: tools take an optional
project_idparam and fall back to the CLI's current project. There is noprojects usetool; the default is whatevertraceway projects useset on the host machine. query_metricsrejectsp50/p95/p99aggregations outright (the CLI accepts them but the server silently computes avg). Latency percentiles come fromlist_endpoints.- No confirmation gate: write tools execute immediately, so only call them on an explicit user request. The MCP client's own tool-approval flow is the confirmation.
--content-fileon the post-mortem commands becomes thecontent_mdstring param. - Login, logout, and profile management are CLI-only (
traceway loginin a terminal); the MCP server uses the credentials stored by the CLI. - Shell-piping recipes (
jqfilters) in the knowledge files describe CLI usage; with MCP tools, read the same fields from the structured tool result instead.