getsentry / getsentry/sentry-mcp
search_events returns HTTP 404 while search_issues + search_issue_events succeed (mcp.sentry.dev hosted)
- Dominant language
- TypeScript
- Stars
- 853
- Forks
- 144
- Avg merge
- 19h 25m
- Merged PRs (30d)
- 32
Description
## Environment
- **Server:** Hosted remote MCP server at `https://mcp.sentry.dev/mcp`
- **Client:** Claude Code (OAuth)
- **Organization:** `info618` (region: `us.sentry.io`)
- **Project:** `b2m-hub`
- **Date observed:** 2026-05-07
## Symptom
`search_events` consistently returns `HTTP 404: API request failed: Not Found` regardless of parameters. The other AI-agent-backed tools (`search_issues`, `search_issue_events`) work fine against the same org/project, which makes this distinct from prior outages where the embedded agent failed for all three together.
## Affected tool
- ❌ `search_events` — HTTP 404 from Sentry REST API on every call
## Working tools (verified in same session, same org)
- ✅ `whoami`
- ✅ `find_organizations` (returns `info618`, regionUrl `https://us.sentry.io`)
- ✅ `find_projects` (returns `b2m-hub`)
- ✅ `search_issues` — returns 17 unresolved issues with full metadata
- ✅ `search_issue_events` — returns full event list with stacktrace, tags, geo
- ✅ `get_sentry_resource` (`resourceType=issue`) — returns full issue + first event detail
## Steps to reproduce
All four variations below return the same 404:
1. **Explicit dataset + Sentry-syntax query, no project**
```
search_events(
organizationSlug='info618',
regionUrl='https://us.sentry.io',
dataset='errors',
query='level:error',
statsPeriod='24h',
limit=5
)
```
2. **Explicit dataset + project**
```
search_events(
organizationSlug='info618',
regionUrl='https://us.sentry.io',
projectSlug='b2m-hub',
dataset='errors',
query='level:error',
statsPeriod='24h',
limit=5
)
```
3. **Natural-language query**
```
search_events(
organizationSlug='info618',
regionUrl='https://us.sentry.io',
query='how many errors today',
limit=5,
includeExplanation=true
)
```
4. **No regionUrl**
```
search_events(
organizationSlug='info618',
dataset='errors',
query='level:error',
statsPeriod='24h',
limit=5
)
```
## Expected
Either an aggregate count (for natural-language) or a list of error events (for explicit dataset query), as documented in the tool description.
## Actual
```
**Input Error**
There was an HTTP 404 error with your request to the Sentry API.
API error (404): API request failed: Not Found
. Please verify the parameters are correct.
You may be able to resolve the issue by addressing the concern and trying again.
```
## Why this is distinct from prior incidents
| Issue | Symptom | Our case |
|---|---|---|
| #779 | OpenAI key invalidated → all 3 search tools fail | search_issues works for us |
| #776 | "No object generated: could not parse the response" | We get 404 from Sentry REST, not a parse error |
| #748/#749 | Cloudflare→OpenAI region block → "Country, region, or territory not supported" | search_issues uses the same agent layer and works for us |
| #553 | Agent generates incorrect aggregation queries | We get 404 even on simple `level:error` Sentry-syntax queries that should bypass aggregation logic |
The fact that `search_issues` succeeds rules out an agent-layer outage. The 404 is coming from the downstream Sentry REST call (likely `/api/0/organizations/info618/events/`), suggesting the URL `search_events` constructs is invalid for this org while the URL `search_issues` constructs is valid.
## Adjacent unrelated bug observed
Earlier in the same session, `search_docs` returned `HTTP 400: Invalid request` on simple queries (e.g., `query='Discover events API plan tier requirement'`, `maxResults=3`). Probably worth a separate triage.
## Notes
- Sentry org has 17 unresolved issues right now — plenty of error data exists for `search_events` to return.
- `search_issue_events` against `B2M-HUB-4R` returned 3 detailed events including stacktraces — confirming the events dataset itself is reachable for this org.
- Repro is 100% — 4 distinct call shapes, all 404.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.