Better Stack vs Sentry MCP

Better Stack 46

The server is closed and tools/list needs sign-in, so the tool schemas were not read, and no OpenAPI file was found for the REST APIs in the docs, the GitHub organisation or a search. Partial credit for a typed contract that exists but isn't public (10 of 25). No llms.txt on betterstack.com or its docs, and no Markdown twins (0). The MCP docs page gives one line per tool and says when some apply, such as Reopen incident within 24 hours and the instruction tools to call before Execute query. The REST reference pages describe each endpoint in a sentence (12 of 20). Inputs are typed only in the REST docs, with parameter tables, monitor types and status values listed in prose, and no schema to validate against (7 of 15). Every REST page carries a curl example and a 200 body; Create monitor documents a 422 with errors keyed by field, the Usage API a 401 and a 429, and no error code list exists (9 of 15). The APIs are versioned v2 in the path. The changelog is a product feed of 19 entries since March 2025 with the MCP server among them, with no API changelog and no deprecation notes (8 of 15). 10 + 0 + 12 + 7 + 9 + 8 = 46.

Sentry MCP 87

Every tool has a Zod input schema published as JSON Schema in toolDefinitions.json (25). llms.txt at docs.sentry.io and linked from mcp.sentry.dev (10). Descriptions are long and structured, with "Use this tool when", <examples> and <hints> blocks, and some say when not to use the tool, for example add_issue_note warns against secrets and unapproved content (18). Inputs carry required lists, minLength and maxLength, URI formats and nullable unions (13). Examples sit inside the descriptions, and errors map to typed classes with recovery hints (13). GitHub releases and tags. The release guide says to update a CHANGELOG.md the repository doesn't have (8).

Better Stack 55

148 tools are named on the docs page, more than 30 (5). X-MCP-Tools-Only and X-MCP-Tools-Except narrow the list to a subset or a read-only set, documented for that purpose, and the server has no default toolset (add 10, so 15). REST lists take page and per_page up to 250 with first, last, prev and next links, and filters by url, pronounceable_name and team_name. Plan query windows splits a long telemetry query into windows, and Execute query takes SQL with its own LIMIT. No field selection (16 of 20). 422 bodies name the field that failed and 401 gives a message, but no error code list exists and MCP errors are not documented (10 of 20). No idempotency keys. Upsert catalog records avoids duplicates, Reopen incident is limited to 24 hours, and the annotations on the tools couldn't be read (6 of 20). Get workspace style defaults are not described. Two Terraform providers and logging clients exist, but no official API SDK in any language, and the Claude plugin's 16 skills carry the query recipes (8 of 15). 15 + 16 + 10 + 6 + 8 = 55.

Sentry MCP 78

The hosted server deploys from main after CI passes (deploy.yml), and main's default tools/list now has 14 tools, find_organizations, find_projects, update_issue, analyze_issue_with_seer, search_issues, get_sentry_resource, the two catalogue tools and six search tools (search_errors, search_logs, search_traces, search_metrics, search_profiles, search_replays) where 0.42.0 had one search_events. Their generated definitions run to about 38,000 characters (name, description and input schema), against about 23,600 for the nine in 0.42.0, the npm latest. A lint rule added on 28 October (#1700) now locks the list. 11 to 30 tools, less 4 for size as the nine were (11). The other 63 tools load on demand through search_sentry_tools and execute_sentry_tool, skills narrow the set, and org or project scoping in the URL hides discovery tools (plus 7). Limits on searches and org and project constraints (17). Errors come back as typed, actionable messages, with 404s telling the agent to check the org, project or id (18). Catalogue tools set readOnlyHint, destructiveHint and idempotentHint, but every catalogue tool runs through execute_sentry_tool, still marked destructiveHint: true as a whole, and issue #1254 (open since 14 August per the 1 October check, not re-read on 1 November) says this breaks client allowlists and approval prompts (15). Few required parameters, npm only (10). 18+17+18+15+10 = 78.

Better Stack 60

OAuth sign-in with a Full access level chosen at consent, or API tokens that are global or scoped to one team with Read & write or read access, minted and deleted in Settings. No scopes finer than read and write, no expiry and no PKCE or dynamic registration documented (24 of 30). Read-level tokens and consent, the two header allowlists, and the rule that comments can only be changed by their author or a token. No confirmation step before Remove monitor, Remove dashboard, Resolve incident or Remove team member (12 of 20). Logs, spans, errors and replays are untrusted text returned to the model, and no injection guidance was found in the MCP docs (2 of 15). Audit logs record every API token request with actor_type api_token and the token's name, never its value, and can be streamed to a Telemetry source or any URL, though the pricing page lists audit logs as a $250 a month add-on (12 of 15). SOC 2 Type 2 with the report on request, third-party penetration tests, security@betterstack.com for disclosures. No security.txt (404), no bug bounty and no disclosure policy page (10 of 20). 24 + 12 + 2 + 12 + 10 = 60.

Sentry MCP 75

The hosted server runs its own OAuth provider with dynamic or metadata-document client registration, and the user picks skills at consent. The upstream Sentry token it holds always asks for org:read project:write team:write event:write alerts:write, so the narrowing happens in the MCP layer only. A 21 October commit (#1561, main) records the scopes Sentry granted but still says the upstream request asks for every scope (25). Inspect and Seer, both read-only, are the default skills. Triage and project management are opt-in, ?skills= narrows a direct token, and org or project constraints limit scope. No confirmation step beyond annotations (16). Contributor docs forbid passing untrusted error text to agents and list input validation and region-URL checks, but event messages, breadcrumbs and user reports still reach the model unmarked (9). Sentry's audit log records user actions, and we didn't confirm MCP calls appear there (8). SOC 2 Type I and II, ISO 27001, HIPAA attestation and a disclosure programme at security@sentry.io. security.txt names a private HackerOne bug bounty that isn't taking new invitations. No advisories for the MCP server were found in the clone; GitHub's advisory page wasn't readable on 1 November. The same evidence with a private bounty scored 17 on nylas-calendar and pplx-decider (17).

Better Stack 50

Read with the hosted lines and scored on the remote MCP server at mcp.betterstack.com, which fronts the same service as the Incidents and Telemetry APIs. status.betterstack.com runs on Better Stack's own status page product with three components and a 90-day history (20). Between 5 August and 3 November 2026 it records delayed Telemetry data processing in the Europe region from 3 to 6 September in entries of 6 hours 19 minutes, 10 hours 10 minutes, 19 hours 13 minutes, 21 hours 45 minutes and 1 day, a 9-minute Telemetry degradation on 20 August, 39 minutes of wrong downtime notifications on 21 October and 8 minutes of dashboard unavailability on 2 November, every one labelled degraded, plus two migrations labelled maintenance (11 hours 13 minutes on 9 August and 17 hours). About four days of delayed log processing in one region is more than trivial and less than an outage of the API, so between minor-only and one major (15). No rate limit numbers were found in the API or MCP docs (0). The Usage API documents a 429 with the instruction to wait before retrying, with no Retry-After, no backoff guidance and no idempotency keys. Upsert catalog records is documented as not creating duplicates when repeated (5 of 15). No SLA. The terms supply the service as is, and the enterprise page lists SLA reporting as a feature (0). The MCP server was announced on 15 September 2025 and carries no beta label (10). 20 + 15 + 0 + 5 + 0 + 10 = 50.

Sentry MCP 36

Scored as a hosted MCP server, since mcp.sentry.dev is the default install. status.sentry.io is an Atlassian Statuspage with API, Dashboard and per-region ingestion components, but no MCP component (15). The history pages rendered on 1 November (the 1 October check could read only the front page). The 90 days from 3 August hold 23 incidents (5 critical, 10 major, 8 minor by Sentry's labels, maintenance excluded). Most name one region (US, EU or DE) or one product such as span ingestion or the dashboard, and count as partial. Four name no region or component limit. On 27 August "sentry.io is not available" ran 09:24 to 15:41 UTC and "affects the availability of our product and API". On 1 September requests timed out with 504s, 08:01 to 11:21 UTC. On 10 September 500 errors rose during a control silo migration, 15:25 to 16:26 UTC. On 20 October users had "interruptions using our UI and API", 18:20 to 21:38 UTC. Several majors (0). The hosted server's limits, 60 calls a minute per user and 300 per IP on /mcp, are in the public Wrangler config, not in user docs (8). Upstream 429s become a typed rate-limit error carrying retry_after when Sentry sends it, documented only in the contributor docs, and since 27 October (#1673, on main) rate-limited GETs are retried from Sentry's reset headers (8). No SLA on the pricing page (0). Versions are 0.x and ?experimental=1 variants can change between releases, with no GA statement (5). 15+0+8+8+0+5 = 36.

Better Stack 69

Closed service under Terms of Use (14 February 2025) that name Better Stack, Inc., a Delaware corporation, under New York law. The plugins and Terraform providers are MIT (15 of 30). The privacy policy (25 November 2022) gives no retention periods and says data is processed primarily in the EU, while the DPA (25 February 2025) names the United States and other locations and promises return or destruction on termination. Telemetry retention is a plan term (3 days free, 30 days in bundles). OpenAI is a subprocessor and nothing says whether customer data is used for model training (14 of 30). No deprecation policy. The terms allow changes at Better Stack's discretion with continued use as acceptance, and the changelog carries no deprecation notices (3 of 20). Schedule A of the DPA lists nine subprocessors with locations, the security page says data is stored in EU regions by default with custom locations for enterprise, and no notice period for new subprocessors is given (16 of 20). 15 + 14 + 3 + 16 = 48.

Sentry MCP 81

FSL-1.1-Apache-2.0, source-available with a two-year conversion to Apache-2.0, not an OSI licence today (20). Sentry publishes a privacy policy, DPA and a subprocessor list updated 1 June 2026 (per the 1 October check, not re-read on 1 November) that names Anthropic and OpenAI as AI subprocessors and Cloudflare, AWS and GCP in the EU and US. TELEMETRY.md shows the hosted server records tool arguments and tool results in Sentry's own monitoring projects, and mcp.sentry.dev still didn't say so on 1 November. A tool added on main on 30 October (#1755), submit_feedback_to_sentry_team, lets an agent send friction reports to Sentry (22). The docs skill is marked deprecated in its own description and tools have been removed with commit notes, but there's no deprecation policy or dates (8). Subprocessors and their locations are disclosed. The stdio package sends telemetry only when a DSN is set (20).

Better Stack 71

Read as a closed service. The Claude plugin reached 1.3.12 on 2 November 2026 and the changelog's newest entry is 29 October 2026 (30). Changelog entries on 10 August, 9 September, 1 October, 15 October and 29 October 2026 (20). Public changelog, hello@betterstack.com with a stated reply within a few hours, and a community site. GitHub issues on the plugin repositories were not read (10 of 15). The MCP server is not in the official MCP registry under a verified namespace, the only entries matching the name being third-party scrapers, and there is no official API SDK. The Claude, Cursor, Codex and Antigravity plugins and two Terraform providers are current (5 of 15). The Claude plugin repository is MIT with 12 commit days since June 2026, and we didn't read its CI (6 of 10). 30 + 20 + 10 + 5 + 6 = 71.

Sentry MCP 80

0.42.0 on 2026-09-25 is still the newest MCP release, the registry latest, 37 days before 1 November (20). The same repository's CLI tagged cli@0.46.0 on 1 October and cli@0.47.0 on 5 October, but neither tag contains the MCP server. The sentry mcp command arrived on main later and is in no CLI release yet, so the CLI tags don't count for this listing. 0.38.0 (26 August), 0.39.0 (27 August), 0.40.0 (24 September), 0.41.0 and 0.42.0 (25 September), five inside the 90 days from 3 August (20). 69 open issues and 33 open pull requests, with unlabelled enhancement requests from July and August and #1226 (hosted AI search failing with "Feature Unavailable", 4 August) open, per the 1 October check, not re-read on 1 November (15). Listed in the official MCP registry as io.github.getsentry/sentry-mcp, 0.42.0 published 25 September 14:49 UTC and marked latest (15). Tests, smoke tests and a token-cost check run in CI on every pull request (10). 20+20+15+15+10 = 80.