# Changelog ## November 13, 2026 - `execute_code_write` (`code-exec` and `core` toolsets) is now available to everyone. It runs scripts that can create, update, and delete Datadog resources, while `execute_code` stays read-only. Your existing MCP write permissions and per-API permissions still apply, and you may need to reauthorize an existing connection. Creating or updating dashboards and notebooks isn't supported through it; use `upsert_datadog_dashboard` and the notebook tools instead. - `search_datadog_dashboards` (`core` toolset) now reports how many dashboards matched and takes a `limit` of up to 200 results. `start_at` is deprecated. - `get_datadog_incident` (`core` toolset) now includes meeting transcripts from the incident timeline. ## November 11, 2026 - `execute_code` (`code-exec` toolset) scripts can now run DDSQL queries, including queries over Error Tracking error events with `dd.errors`. - `get_synthetics_tests` and `get_global_variables` (`synthetics` toolset) now say whether there are more results and how to get the next page. ## November 10, 2026 - `apm_get_service_health`, `apm_latency_bottleneck_summary`, `apm_search_recommendations`, `apm_get_recommendation`, and `apm_discover_span_tags` (`apm` toolset) are now generally available. - `synthetics_test_wizard` (`synthetics` toolset) now supports multistep API tests and JSONPath, XPath, JSON Schema, and JavaScript assertions. - `upsert_datadog_dashboard` (`core` toolset) now accepts `default_timeframe` to set or clear the time range a dashboard opens with. ## November 6, 2026 - New `search_cost_recommendations` and `aggregate_cost_recommendations` tools (`cost` toolset) for finding cloud cost recommendations and summarizing them, such as total potential savings by resource type. ## November 5, 2026 - The monthly MCP tool-call limit has been removed. Short-term burst limits still apply. - `ddsql_validate_query` (`ddsql` toolset) is now available to everyone. It checks a DDSQL query for errors without running it. - `search_datadog_metrics` (`core` toolset) now takes its search term as `query`, like the other search tools. `name_filter` still works for now but is deprecated. ## November 4, 2026 - `search_datadog_spans` (`core` toolset) now returns fields under the same names you use in queries (`span_id`, `trace_id`, `resource_name`, …), so values can be copied straight into a filter. - `explore_profiling_rum_flamegraph` (`profiling` toolset) now supports RUM operations and mobile apps on iOS and Android, including app launch vitals. ## November 3, 2026 - The `investigator` toolset is now generally available. It lets you start a Bits AI investigation (generally, or for a specific incident), search existing investigations, and fetch their results. - `search_datadog_spans` and `aggregate_spans` (`core` toolset) now handle attribute filters written without the `@` prefix, like `http.status_code:500`. Previously these silently matched nothing. - `search_datadog_logs` (`core` toolset) now reports log pattern counts as estimated totals rather than counts within the sampled logs. - The profiling exploration tools (`profiling` toolset) are now all scoped the same way, through `queryString` or trace context. The `runtimeId` and `profileContext` arguments have been removed. To target one profile, put `profile-id:` in `queryString`. ## November 2, 2026 - `aggregate_spans` (`core` toolset) now accepts `duration` as a compute field instead of rejecting it and asking for `@duration`. ## October 30, 2026 - `get_datadog_trace` (`core` toolset) now reports span start and end times to the microsecond. Previously they were rounded to the second, which made it impossible to order overlapping spans. - New `move_datadog_spreadsheet_tab` tool (`sheets` toolset) for reordering tabs. `get_datadog_spreadsheet` now also returns the tab order. - `execute_code` (`code-exec` toolset) now says which value was wrong, and what's allowed, when a Datadog API request fails validation. ## October 29, 2026 - `get_datadog_dashboard` and `upsert_datadog_dashboard` (`core` toolset) now return a `revision_id`. Pass it back to `upsert_datadog_dashboard` as `base_revision_id`, and the update is rejected if someone else changed the dashboard since you read it, instead of silently removing widgets they added. ## October 27, 2026 - RUM tools (`rum` toolset) now include links to the matching Datadog page in many more responses. ## October 26, 2026 - `run_network_path` (`networks` toolset) now returns a link to the Network Path result. - `analyze_datadog_logs` (`core` toolset) now suggests a supported alternative when a query uses an SQL function that logs don't support, such as `SUBSTR` or `REGEXP_EXTRACT`. ## October 22, 2026 - RUM retention filter tools (`rum` toolset) now see and manage exclusion filters. Previously, trying to delete an exclusion filter reported success without deleting anything. - `upsert_datadog_spreadsheet` (`sheets` toolset) now supports conditional formatting on flex sheets. ## October 21, 2026 - `search_datadog_metrics` (`core` toolset) accepts `any` for `percentiles_filter`, for clients that always send every optional argument (such as OpenAI's strict mode). Previously those clients could only hide metrics without percentiles. ## October 19, 2026 - The `derived-metrics` toolset is now generally available. Its `create_derived_metric`, `update_derived_metric`, and `delete_derived_metric` tools manage derived metrics. - `aggregate_spans` (`core` toolset) now accepts a sort placed inside a compute, instead of rejecting the request. ## October 16, 2026 - `aggregate_rum_events` (`rum` toolset) now says when RUM isn't enabled for your organization, matching `search_datadog_rum_events`. Previously it returned empty results. - `search_datadog_rum_events` (`core` and `rum` toolsets) now accepts `sort`, like the other search tools. - `search_datadog_incidents` (`core` toolset) now returns every ongoing incident when called without a query, including `stable` ones and any custom live statuses. Previously it only returned `active` incidents. ## October 15, 2026 - Relative time ranges now accept year units, so values like `now-1y` work in every tool that takes a relative time. Previously they were rejected. - `get_datadog_spreadsheet_table_data` (`sheets` toolset) has been removed. Use `get_datadog_spreadsheet_tab_data` instead, which covers the same data. ## October 13, 2026 - `execute_code` (`code-exec` toolset) is now marked as read-only, matching what it can actually do. Clients that treat write-capable tools differently, such as by asking for approval first, now classify it correctly. - `update-feature-flag-environment` (`feature-flags` toolset) can now enable or disable a flag in production environments, subject to your permissions and approval policy. Previously you had to switch to the Datadog UI for this. Default and override variant changes are still limited to non-production environments. ## October 9, 2026 - `search_rum_operations`, `get_rum_operation_summary`, and `get_rum_operation_insights` (`rum` toolset) no longer accept `lookback`. Use `from` and `to` instead, like the other RUM tools. The default window is the last 7 days. ## October 8, 2026 - `search_datadog_rum_events` (`core` and `rum` toolsets) now pages through results with a `cursor`, which replaces `start_at`. Previously, paging past the first 1,000 events returned an empty result that was reported as complete. ## October 7, 2026 - RUM events tools (`rum` toolset) now include a RUM Explorer link, and `search_datadog_rum_events` results include a Session Replay link for events that have a replay. - `execute_code` (`code-exec` toolset) can now read Datadog hourly usage and billable usage summaries, and look up which teams a user belongs to. ## October 6, 2026 - `create_datadog_logpoint` (`live-debugger` toolset) can now target every environment at once with `environment: "*"`. Previously this failed. ## October 5, 2026 - `get_datadog_metric_context` (`core` toolset) now returns JSON instead of YAML, matching `search_datadog_metrics` and `get_datadog_metric`. ## September 30, 2026 - New `create-feature-flag-variant` and `update-feature-flag-variant` tools (`feature-flags` toolset) for adding variants to an existing flag and renaming or changing their values. - `explore_profiling_rum_flamegraph` (`profiling` toolset) now requires `rumQuery` and rejects blank or match-all queries with a clear message. Calls that left it out used to fail with an unhelpful error. - `get_datadog_metric` (`core` toolset) cloud cost queries on `datadog.cost.amortized` now work over OAuth connections. If you're missing the billing or usage permission they need, the response now names the missing permission instead of giving a generic no-data warning. - The events tools (`core` toolset) now tell you which argument to fix, with query syntax tips, when a request is rejected. ## September 29, 2026 - Fixed `get_datadog_database_explain_plans` (`dbm` toolset) not returning plans for SQL Server stored procedures. - `search_datadog_rum_events` (`core` and `rum` toolsets) now accepts a `limit` parameter. ## September 28, 2026 - The `visualizations` toolset is now generally available. It includes `get_widget` and `visualize_tabular_data`. ## September 25, 2026 - `benchmark_datadog_database_query` (`dbm` toolset) now accepts a `table_row_counts` parameter, so you can benchmark against realistic table sizes instead of the default 10,000 rows per table. ## September 24, 2026 - Software delivery tools (`software-delivery` toolset) now accept date-only time bounds and a `limit` alias on search, reject reversed or empty time windows up front, and return actionable errors instead of opaque ones. Read-only calls are also retried automatically on transient upstream failures. - **Breaking change for `execute_code` (`code-exec` toolset):** the `traces` extension has been removed from `@datadog/code-exec`. Scripts that imported it should use the SDK's APM trace API instead, which covers both full and pruned traces. Note the response shape differs from the old extension's. ## September 23, 2026 - Fixed `search_datadog_spreadsheets` (`sheets` toolset) returning the wrong page of results when `start_at` landed on a page boundary — some spreadsheets were unreachable entirely. - `get_datadog_k8s_manifest` (`kubernetes` toolset) now explains what to do when a manifest fetch is rate limited, instead of surfacing a bare error. - `execute_code` (`code-exec` toolset) can now fetch traces — full or pruned — including from archived sessions, and read a much wider set of Agent Observability data. ## September 22, 2026 - `ddsql_upsert_saved_query` (`ddsql` toolset) now respects your organization's MCP write protection setting. Previously saved queries could be created or updated even with MCP writes disabled. ## September 16, 2026 - `analyse_datadog_k8s_rollout` (`kubernetes` toolset) is now available to everyone with the toolset enabled. It returns a Deployment rollout in a single call: status, progress, ETA or duration, the ReplicaSet split by revision, and before/after impact on service performance, resource usage, and log volume. ## September 15, 2026 - The RUM operation tools (`rum` toolset) — `search_rum_operations`, `create_rum_operation`, `update_rum_operation`, `delete_rum_operation`, `get_rum_operation_summary`, and `get_rum_operation_insights` — are now available to everyone with the toolset enabled. ## September 11, 2026 - `create_datadog_security_findings_automation_rule` (`security` toolset) now supports Linear as a ticket creation target, alongside Jira and Case Management. Linear-specific fields go in `fields` (`linear_project_id`, `linear_label_ids`). - `analyze_cloud_network_monitoring` (`networks` toolset) now rejects wildcard tag values and unrecognized `group_by` tags with a clear error, instead of silently returning results that don't match what you asked for. ## September 9, 2026 - The `code-exec` toolset — `execute_code` and `search_datadog_sdk`, for running sandboxed TypeScript against the Datadog APIs — is now generally available and visible to clients. - Fixed `analyze_cloud_network_monitoring` (`networks` toolset) using the wrong facet names for client and server availability zone, which made AZ-based filters and group-bys come back empty. ## September 8, 2026 - Datadog MCP tools now load in Vertex AI and Gemini clients, which previously failed to load them at all because of how optional parameters were declared. - `aggregate_events` and `aggregate_spans` (`core` toolset) and `aggregate_rum_events` (`rum` toolset) now accept `P98`, which was previously rejected even though the underlying API supports it. - `get_datadog_database_schemas` (`dbm` toolset) now returns indexes, foreign keys, and partition counts for each database instance. These fields were part of the tool's contract but always came back empty. ## September 7, 2026 - Agentic onboarding (`onboarding` toolset) is no longer restricted to a fixed preview list of language and deployment combinations — any combination the onboarding strategies support is now available. ## September 4, 2026 - `get_datadog_database_explain_plans` (`dbm` toolset) now accepts a `plan_signature` filter, so you can fetch a specific plan directly instead of paging through every plan for a query. ## September 3, 2026 - When you connect over OAuth with read-only access, the server now says so up front: write-capable tools are omitted, and the instructions explain how to reauthorize with full access if a task needs one. - Fixed `analyze_cloud_network_monitoring` (`networks` toolset) reporting elevated retransmits as outright network loss. - A single oversized response no longer kills an `execute_code` (`code-exec` toolset) run. Results over the size limit now come back as an explicit `result_too_large` failure, and other API calls in the same script still return their results. - Archived timeseries aggregations in `execute_code` (`code-exec` toolset) no longer require an explicit `interval`. Omitting it now gives you the documented default of 150 points, matching the live logs, spans, and RUM APIs. ## September 2, 2026 - `ddsql_run_query` (`ddsql` toolset) now accepts optional `from` and `to` parameters (ISO 8601, Unix seconds or milliseconds, or relative forms like `now-1h`). Without them, queries keep the previous one-hour default window. - `get_datadog_database_recommendations` (`dbm` toolset) now includes MongoDB recommendations for queries that scan far more documents than they return. - `optimize_datadog_database_query` (`dbm` toolset) now uses the LLM-backed optimization path, so it returns rewritten query suggestions rather than only observations. - Archived aggregation queries in `execute_code` (`code-exec` toolset) now support nested group-bys, keeping the sort and limit you set at each level instead of collapsing them into a single global sort and limit. ## September 1, 2026 - `analyze_cloud_network_monitoring` (`networks` toolset) now includes a Network Analytics deep link with each result, and no longer lets the `Missing`/`N/A` bucket crowd real groups out of the response. ## August 31, 2026 - `get_datadog_k8s_manifest` (`kubernetes` toolset) now accepts an `exclude_fields` parameter, so you can drop specific fields while keeping the rest of the manifest intact. Previously `concise: true` was the only way to trim a manifest, and it also stripped empty-but-required fields, which could make the output an invalid spec. ## August 28, 2026 - The default monthly MCP tool call quota has doubled, from 50,000 to 100,000 calls per organization. ## August 26, 2026 - New `explore_profiling_rum_flamegraph` tool (`profiling` toolset) returns a flamegraph for profiles scoped to RUM events matching a metric such as INP or LCP, resolving the matching RUM events server-side. Previously this meant fetching RUM event IDs yourself and passing them to `explore_profiling_flame_graph`. ## August 20, 2026 - `search_datadog_spans` (`core` toolset) gained a `limit` parameter (1–1000), cursor-based pagination, and a `custom_attributes` parameter for trimming returned attributes to the ones you care about. `start_at` has been removed in favor of `cursor`. ## August 19, 2026 - `execute_code` (`code-exec` toolset) can now run timeseries aggregations against archived logs, spans, and RUM data, not just scalar ones. ## August 14, 2026 - `execute_code` (`code-exec` toolset) gained read access to Network Path test runs (live and archived), Cloud Cost Management, and Cloud Network Monitoring connection and DNS aggregates. Writes remain blocked. ## August 11, 2026 - `get_datadog_trace` and `aggregate_spans` (`core` toolset) now include a UI deep link in more response cases — including trace-not-found results and traces without a root span — for live (non-archive) queries. ## August 10, 2026 - `list-stale-feature-flags` and `clean-up-flag` (`feature-flags` toolset) now honor an org-level staleness-window override instead of always using a fixed 30-day cutoff when determining which flags are stale. - `get_autonomous_system_status` (`networks` toolset) now automatically pulls in matching Network Path test run evidence when reporting a degraded autonomous system. ## August 7, 2026 - `get_datadog_metric` (`core` toolset) now includes a link to the Metrics Explorer for scalar query results, matching the link already provided for timeseries results. - `create-feature-flag` and the allocation-sync tools (`feature-flags` toolset) now accept a future `scheduled_start` for CANARY/FEATURE_GATE allocation rollouts — previously this was rejected. ## August 6, 2026 - New `list_autonomous_system_statuses` tool (`networks` toolset) reports which autonomous systems are currently degraded across your org, without needing to name a specific AS. ## August 4, 2026 - Serverless onboarding (`onboarding` toolset) now supports Azure Container Apps as a compute target, and adds PHP as a supported runtime for GCP Cloud Run, Cloud Run functions, and Azure Container Apps. - `search_pr_insights` (`software-delivery` toolset) now documents that an `expected` product status means results are still pending, not that nothing was found. ## July 31, 2026 - New `update_data_observability_recommendation_status` tool (`data-observability` toolset) lets you mark a Data Observability recommendation as applied or dismissed. `list_data_observability_recommendations` gained a `tag_query` argument for filtering by tags, and `get_data_observability_recommendation` now returns the recommendation's tags. - `search_datadog_k8s_resources`, `describe_datadog_k8s_resource`, and `get_datadog_k8s_manifest` (`kubernetes` toolset) are more forgiving on the `kind` argument — accepting case variants, plurals, and kubectl shorthand (`pods`, `svc`, `deploy`), plus clearer errors listing valid kinds. - `describe_datadog_k8s_resource` (`kubernetes` toolset) now includes container statuses, init container statuses, pod conditions, and the first failing condition's message when describing a pod. ## July 30, 2026 - Serverless onboarding (`onboarding` toolset) now provides dedicated OpenTelemetry setup guidance for GCP Cloud Run instead of the standard Datadog Agent flow. ## July 29, 2026 - `update_datadog_security_signals_triage` (`security` toolset) no longer fails with a permission-denied error when triaging by `filter_query` — it now correctly requires signals read + write permissions instead of write alone. - `create_datadog_form`, `update_datadog_form`, `clone_datadog_form`, `get_datadog_form`, `search_datadog_forms`, and `publish_datadog_form` (`forms` toolset) now return a `form_url` field linking directly to the form in the Datadog UI. ## July 28, 2026 - `upsert_datadog_spreadsheet` (`sheets` toolset) now drops empty-content cells before saving, so creating a spreadsheet — or adding a new sheet to an existing one — no longer fails when the definition contains blank cells. - `search_datadog_k8s_resources` (`kubernetes` toolset) accepts a new `include_deep_link` option that adds a link to the matching query in the Orchestration Explorer UI. ## July 24, 2026 - `upsert_datadog_dashboard` and `get_datadog_dashboard` (`dashboards`, `core` toolsets) now support additional widget types: Product Analytics cohort, retention, and funnel, plus Sankey and wildcard widgets. - New Data Observability monitor annotation tools (`data-observability` toolset) for inspecting, creating/updating, and deleting annotations on data quality monitors, plus entity-aware monitor coverage. - `list_datadog_database_optimizations` (`dbm` toolset) now supports pagination, filtering by optimization type, and sorting by estimated improvement. - `get_datadog_spreadsheet_tab_data` (`sheets` toolset) now reads pivot tabs, in addition to table tabs. - `search_dora_events` (`software-delivery` toolset) now returns the full event object (as YAML), so newly added fields surface automatically. ## July 23, 2026 - `get_datadog_dashboard` and `upsert_datadog_dashboard` (`dashboards`, `core` toolsets) now support dashboard tabs, including per-tab widget layout. - Redesigned DORA aggregation: `aggregate_dora_events` (`software-delivery` toolset) now takes composable `queries` and `formulas` (like `get_datadog_metric`) instead of fixed metric types. A new `get_dora_fields` tool discovers the available measures, facets, and aggregations per DORA index. ## July 21, 2026 - `cost_recommendations` (`cost` toolset) can now be sorted by savings, risk, or level of effort. ## July 20, 2026 - `generate_monitor_message` (`alerting` toolset): the `type` parameter is now required, and unsupported monitor types return a clear error instead of failing server-side. - Audit Trail tools (`audit-trail` toolset) now include an `audit_trail_explorer_url` in responses — a deep link into the Audit Trail Explorer scoped to the query and time window you just ran. - Dashboard tools (`dashboards`, `core` toolsets) now read and write more template variable fields: group-by variables (`type`), a dynamic `available_values_query`, and per-data-source `data_source_mappings`. - `get_datadog_spreadsheet_table_data` (`sheets` toolset) is deprecated in favor of `get_datadog_spreadsheet_tab_data`. ## July 16, 2026 - New feature flag onboarding tools (`feature-flags` toolset) that guide you through setting up your first flag: detecting your stack, creating a non-production flag, wiring up the SDK, and verifying it end to end. ## July 15, 2026 - `cost_recommendations` (`cost` toolset) now returns a risk level and a level-of-effort estimate for each recommendation. - `get_datadog_security_detection_rules_schema` (`security` toolset) now includes a `rule_search_facets` section describing the facets you can search detection rules by. ## July 14, 2026 - `search_dora_deployments` is renamed to `search_dora_events` (`software-delivery` toolset) and can now retrieve pull requests in addition to deployments. ## July 13, 2026 - `get_synthetics_tests` (`synthetics` toolset) now handles Synthetics test suite public IDs gracefully instead of erroring. ## July 10, 2026 - `get_rum_summary` (`rum` toolset) now reports ANR rate (Android) and hang rate (iOS). - New `get_datadog_spreadsheet_tab_data` tool (`sheets` toolset) for reading spreadsheet tab data. ## July 9, 2026 - New `list_datadog_database_optimizations` tool (`dbm` toolset) that lists query optimization suggestions for your databases. - `get_datadog_dashboard` (`dashboards`, `core` toolsets) now preserves a template variable's `prefix` when it matches the variable name, fixing a round-trip issue where the prefix was lost on read then update. ## July 7, 2026 - New `get_autonomous_system_status` tool (`networks` toolset) that returns Network Path status for a given autonomous system (AS). ## July 6, 2026 - Audit Trail tools (`audit-trail` toolset) now redact large fields (asset diffs and metadata) by default to keep responses compact; pass `detailed_output: true` for the full events. - `get_datadog_error_tracking_issue` (`error-tracking` toolset) now surfaces session replay links and trace IDs for the sampled errors on an issue. - `get_synthetics_tests` (`synthetics` toolset) can now attach failure screenshots to browser-test results via a new `include_failure_screenshots` option. - New Data Observability recommendations tools (`data-observability` toolset): `list_data_observability_recommendations` and `get_data_observability_recommendation`. ## July 3, 2026 - `get_datadog_error_tracking_issue` (`error-tracking` toolset) now surfaces linked GitHub pull requests for an issue, alongside the existing issue details. - `search_datadog_events` (`core` toolset) now accepts `asc` and `desc` as aliases for the `sort` parameter (mapping to `timestamp` and `-timestamp`). These common values were previously rejected. ## July 2, 2026 - `get_datadog_metric` (`core` toolset) now accepts the `formulas` parameter when items are provided as objects (e.g. `{"formula": "query0 + query1"}`), not just as plain strings. Previously these calls were rejected. ## July 1, 2026 - New RUM retention filter write tools (`rum` toolset) for creating, updating, reordering, and deleting retention filters, complementing the existing search tool. Each write requires `confirm: true` since retention filters affect data retention and billing. - `get_datadog_security_detection_rules` (`security` toolset) now supports requesting a subset of fields, so responses can be trimmed to just what you need. - Standardized sorting arguments across several tools: dashboard and notebook search now take `sort` instead of `sort_by` (`dashboards`, `core` toolsets), and `search_datadog_monitors` (`core` toolset) uses `-field` for descending order instead of `field,desc`. ## June 30, 2026 - Fixed `upsert_datadog_spreadsheet` (`sheets` toolset) dropping fields (timeframe, query, schema format) on targeted updates. It now preserves the existing spreadsheet definition instead of wiping unspecified fields. ## June 29, 2026 - The `rum` toolset is now generally available. It includes tools for browsing RUM applications, managing RUM-based metrics, and retrieving RUM insights. `upsert_rum_metric` now also returns a cardinality estimate for the metric you're creating. - `retry_datadog_ci_job` (`software-delivery` toolset) now supports retrying GitLab pipelines, in addition to the existing GitHub Actions support. - Fixed `search_datadog_services` (`core` toolset) mishandling service or team names that contain colons, which were incorrectly parsed as `field:value` filters. ## June 26, 2026 - `explore_profiling_call_graph` (`profiling` toolset): the `service` and `family` top-level parameters have been removed. Scope by service or language family using `filter.query` instead (e.g. `service:my-svc family:java`). ## June 23, 2026 - `get_datadog_notebook` (`core` toolset) now accepts an `include_comments` parameter. When set to `true`, the response includes all comments on the notebook threaded by reply, alongside the notebook cells. ## June 22, 2026 - `manage_datadog_error_tracking_issue_comments` (`error-tracking` toolset) is now available to all users with the `error-tracking` toolset enabled. Previously required the `experimental` toolset. Supports adding, updating, and deleting comments on Error Tracking issues. - `get_datadog_error_tracking_issue` (`error-tracking` toolset) now returns an activity timeline — the most recent comments, state changes, and assignee updates for the issue, matching what's shown in the Datadog UI. - `create_reference_table` (`reference-tables` toolset) now supports creating empty tables without cloud storage. Pass `source: LOCAL_FILE` to create a table with no backing bucket; rows can then be added with `upsert_reference_table_rows` or `append_reference_table_rows`. Previously all tables required an S3, GCS, or Azure source. - Fixed `search_datadog_logs` (`core` toolset) log pattern clustering returning results from live data instead of the archived snapshot when `use_log_patterns: true` was used in an archived log snapshot context. ## June 19, 2026 - New `create_notebook_comments` tool (`core` toolset) for adding comments to cells in a Datadog notebook. - `search_datadog_error_tracking_issues` (`error-tracking` toolset) now surfaces "Did you mean?" suggestions when a search returns zero results, showing similar valid values for facet filters like `service`, `env`, and `version` so agents can correct and retry without an extra round-trip. ## June 17, 2026 - New `get_datadog_security_detection_rules` tool (`security` toolset) for listing and fetching security detection rules. This replaces the old `list_datadog_security_detection_rules` and `get_datadog_security_detection_rule` tools, which have been removed. - `search_datadog_services` (`core` toolset) now returns both the internal service identifier (used in `service:` tag filters) and the human-readable `display_name` from the catalog, so agents can distinguish them correctly. ## June 15, 2026 - Fixed `analyze_datadog_logs` (`core` toolset) incorrectly auto-prepending `@` to column names in `extra_columns`. Column names like `kube_namespace` are now passed through as-is; grouping by tags works correctly without needing to strip or add the prefix manually. - Four `security` toolset AppSec tools have been renamed to match Datadog naming standards: `list_security_trace_passlist` → `get_datadog_security_trace_passlist`, `upsert_security_trace_passlist` → `upsert_datadog_security_trace_passlist`, `delete_security_trace_passlist` → `delete_datadog_security_trace_passlist`, `list_datadog_security_aap_denylist` → `get_datadog_security_aap_denylist`. The old names no longer work. ## June 5, 2026 - New `upsert_reference_table_rows` tool (`reference-tables` toolset) for inserting or updating rows in a reference table by primary key. Tables with composite primary keys are not supported. - Fixed `get_datadog_dashboard` (`dashboards` toolset) rejecting numeric integration dashboard IDs (e.g. `30584`). These IDs now resolve correctly via the integration timeboard and screenboard endpoints. - Fixed `create_datadog_notebook` (`core` toolset) rejecting valid time spans like `2h` and `3h`. These are now accepted; unsupported values (e.g. `7d`) are coerced to the nearest valid span with a note in the response. ## June 2, 2026 - `analyze_datadog_error_tracking_errors` (`error-tracking` toolset) is now available to all users with the `error-tracking` toolset enabled. Previously it required the `experimental` toolset. ## June 1, 2026 - The `workflows` toolset is now available to HIPAA-compliant organizations. ## May 30, 2026 - `manage_datadog_error_tracking_issue_tickets` (`error-tracking` toolset) now supports creating and linking Linear tickets, in addition to Jira tickets and Datadog cases. ## May 29, 2026 - New `retry_datadog_ci_job` tool (`software-delivery` toolset) for queuing retries of failed GitHub Actions CI jobs. - Fixed `search_datadog_logs` (`core` toolset) `extra_fields` matching — field names without a source prefix (e.g. `caller`) now match again and no longer need to be written as `custom.caller` or `attributes.caller`. ## May 27, 2026 - Fixed `search_datadog_metrics` (`core` toolset) returning errors for certain tag filter combinations. ## May 26, 2026 - `update_datadog_security_signals_triage` (`security` toolset) is now available to all organizations with the `security` toolset enabled. - `create_datadog_notebook` and `edit_datadog_notebook` (`core` toolset) now accept a `tags` parameter for setting notebook tags. - Fixed `search_datadog_service_dependencies` (`core` toolset) crashing on services with cyclic dependency graphs. - Fixed `search_datadog_monitors` (`core` toolset) not honoring sort direction when using suffixes like `title,asc` or `title,desc`. ## May 22, 2026 - Error tracking tools (`error-tracking` toolset) now include regression metadata: `get_datadog_error_tracking_issue` includes a `regression` block (with resolved and regressed timestamps and version), and `search_datadog_error_tracking_issues` marks regressed issues with `is_regression: true`. ## May 21, 2026 - Fixed `search_datadog_services` (`core` toolset) not returning results for services whose catalog defines a custom display name. ## May 20, 2026 - New `update_datadog_error_tracking_issue` tool (`error-tracking` toolset) for updating issue state and assignee. - Fixed incident tools (`core` toolset) producing URLs with the wrong base domain for orgs that use a custom Datadog subdomain. - `get_datadog_metric` (`core` toolset) now defaults the `aggregator` to `avg` in scalar mode instead of returning an error when it's omitted. ## May 19, 2026 - Fixed `search_datadog_incidents` (`core` toolset) returning no results when state filters used uppercase letters (e.g. `ACTIVE`). - `get_datadog_flaky_tests` (`software-delivery` toolset) now includes a direct link to each test's history page in Datadog. - The daily tool-call rate limit has been removed; only burst (50 calls/10s) and monthly (50,000 calls/month) limits now apply. - `search_datadog_metrics` (`core` toolset) now silently clamps oversized `max_tokens` values instead of returning an error. - Fixed multi-word fuzzy search in `list_datadog_skills` (`skills` toolset) — searches like "security findings" now return results correctly. - OAuth now accepts the `cursor-dev://` callback URI scheme, allowing Cursor IDE to complete the authorization flow. ## May 18, 2026 - New `run_network_path` tool (`networks` toolset) for triggering on-demand traceroute tests via a Private Action Runner host. ## May 15, 2026 - New `list_reference_table_rows` tool (`reference-tables` toolset) for paginating through rows in a reference table. ## May 14, 2026 - `search_datadog_monitors` (`core` toolset) now accepts an `include_tags` parameter to return monitor tags in results, with wildcard support. ## May 13, 2026 - New `get_datadog_security_filters` tool (`security` toolset) for listing and fetching Cloud SIEM security filters. ## May 12, 2026 - `get_datadog_error_tracking_issue` (`error-tracking` toolset) now includes a `suspect_commit` field when Error Tracking has identified the likely responsible commit. - `search_datadog_error_tracking_issues` (`error-tracking` toolset) now supports filtering by issue state (e.g. `open`, `resolved`, `for_review`). - New `update_datadog_flaky_test_states` tool (`software-delivery` toolset) for quarantining, disabling, or marking flaky tests as fixed. - Fixed `search_datadog_logs` (`core` toolset) returning incomplete error details for log events with both application-level error fields and error-tracking metadata. ## May 11, 2026 - `edit_datadog_notebook` (`core` toolset) now supports renaming a notebook and setting template variables as metadata-only edits, without re-uploading cell content. - `get_datadog_notebook` (`core` toolset) now returns `template_variables` and `tags`. ## May 6, 2026 - Data Streams (`data-streams` toolset) Kafka tools now return a link to the Kafka setup documentation when queried in an environment where DSM or Kafka monitoring hasn't been configured. ## May 5, 2026 - `edit_datadog_notebook` and `create_datadog_notebook` (`core` toolset) now accept the legacy `cells`-as-string format and `id` parameter, restoring compatibility for integrations that haven't yet adopted the April 21 API changes. ## May 4, 2026 - New `analyze_datadog_security_signals` tool (`security` toolset) for AI-powered batch analysis of security signals, now available to all organizations with the `security` toolset enabled. ## April 30, 2026 - Added an `omit_tools` query parameter for disabling specific tools at connection time. - `list_reference_tables` (`reference-tables` toolset) now includes primary key info in schema output. ## April 29, 2026 - New `create_datadog_security_suppression`, `update_datadog_security_suppression`, and `delete_datadog_security_suppression` tools (`security` toolset). - `search_datadog_metrics` (`core` toolset) now returns up to 1,000 results per query (up from a much smaller cap) and produces more reliable results for historical time ranges. - Error tracking tools (`error-tracking` toolset) now return human-readable display names for issue states (e.g., "In Progress" instead of `INPROGRESS`). - OAuth now supports CORS for the `datadoghq.eu` domain so EU-based browser clients can complete the auth flow. ## April 27, 2026 - Log search tools (`core` toolset) now expose `span_id` and `trace_id` in detailed and TSV output, enabling log-to-trace navigation. ## April 24, 2026 - `search_datadog_logs` (`core` toolset) now accepts a `pattern_group_by` parameter for grouping log patterns by a field. - `create_datadog_notebook` (`core` toolset) now tolerates invalid notebook names instead of failing. - Renamed `security_signals_schema` to `datadog_security_signals_schema` (`security` toolset). ## April 23, 2026 - `get_datadog_metric` (`core` toolset) now supports scalar queries and structured query inputs. - New `get_datadog_security_signal` tool (`security` toolset) for fetching individual security signals. - Aggregate tools (`aggregate_rum_events`, `aggregate_spans`, `aggregate_events` in `core` toolset) now surface dropped compute outputs when responses are truncated. ## April 22, 2026 - `append_reference_table_rows` (`reference-tables` toolset) now requires both read and write permissions on reference tables. ## April 21, 2026 - Notebook tools (`core` toolset): the `cells` argument is now an array of structured cells (previously a string), and the notebook ID parameter was renamed from `id` to `notebook_id`. ## April 20, 2026 - `search_datadog_error_tracking_issues` (`error-tracking` toolset) now supports filtering by assignee. ## April 19, 2026 - Aggregate tools (`aggregate_rum_events`, `aggregate_spans`, `aggregate_events` in `core` toolset) now support response flattening, token budgets, and pagination for large result sets. ## April 17, 2026 - The `ddsql` toolset is now generally available. - `search_datadog_logs` (`core` toolset) now accepts a `clustering_pattern_field` argument to cluster logs by a specific field. ## April 16, 2026 - `get_datadog_error_tracking_issue` (`error-tracking` toolset) now includes an `impact` field. - Log search and analysis tools (`core` toolset) now return helpful timeout messages instead of opaque errors. ## April 15, 2026 - `aggregate_rum_events`, `aggregate_spans`, and `aggregate_events` (`core` toolset) now support percentile aggregation. - `create_datadog_notebook` (`core` toolset) can now create private notebooks. - Security signals tools (`security` toolset) now support filtering by `event_tracker_id`. - The `whoami` MCP resource (`core` toolset) now includes `user_uuid`. - Date-only ISO 8601 strings (e.g., `2026-04-15`) are now parsed as midnight UTC. ## April 13, 2026 - The `kubernetes` toolset is now generally available. - The `reference-tables` toolset is now generally available. - Improved `analyze_security_findings` (`security` toolset) tool documentation. ## April 9, 2026 - The `dashboards` toolset is now generally available. - Notebook create/edit tools (`core` toolset) unified around the raw API cell format and now support general widget types. ## April 7, 2026 - `search_datadog_notebooks` (`core` toolset) now accepts `count` and `include_facets` parameters. - `search_datadog_service_dependencies` (`core` toolset) now defaults to `downstream` direction. - Removed unused boolean filter parameters from `search_datadog_metrics` (`core` toolset). ## April 6, 2026 - Fixed `search_datadog_monitors` (`core` toolset) sort parameter validation. ## April 3, 2026 - Fixed a pagination bug in `search_datadog_service_dependencies` (`core` toolset) that could return incomplete results. ## April 2, 2026 - `search_datadog_metrics` (`core` toolset) now searches over the last 2 weeks by default, up from 1 hour. - The OAuth consent screen is now more user-friendly; now shows OAuth client names instead of IDs. ## March 31, 2026 - A new `submit_mcp_feedback` tool (`core` toolset) has been added – agents can use it to tell us about bugs and missing features. ## March 30, 2026 - Aggregate tools (`aggregate_rum_events`, `aggregate_events`, `aggregate_spans`) have been added to the `core` toolset to help agents reason about large numbers of items in aggregate. ## March 27, 2026 - Added `toolsets=all` shorthand that expands to all generally-available toolsets. - Reduced token usage of `core` toolset tool definitions by ~15%. ## March 24, 2026 - Fixed an issue preventing ChatGPT from connecting to the MCP server. ## March 19, 2026 - Added `aggregate_rum_events`, `aggregate_events`, and `aggregate_spans` tools to the new `aggregation` toolset for computing metrics, counts, and groupings over RUM events, events, and spans. (This toolset is experimental — these tools will likely move to other toolsets in the future.) ## March 18, 2026 - `get_datadog_trace` and `search_datadog_spans` (`apm` toolset) now include deep links to the trace view in Datadog, so agents can share direct URLs instead of constructing them. - Fixed a bug where month-based relative time ranges (e.g., `now-2M`) were silently interpreted as minutes. ## March 17, 2026 - Fixed an issue preventing Gemini CLI from connecting to the MCP server - Fixed a bug where `search_datadog_incidents` was not respecting time filters. - Improved reliability of `search_datadog_events` for large result sets. ## March 16, 2026 - Added `whoami` and `toolsets` MCP resources to the `core` toolset, so agents can check which user/org is connected and which toolsets are available. ## March 13, 2026 - Fixed an issue preventing Claude Connectors (Claude web UI and Claude Desktop/Cowork) from connecting to the MCP server - Improved `search_datadog_incidents` tool documentation: clarified commander/responder query syntax, case sensitivity, and wildcard support. ## March 11, 2026 - Replaced `search_datadog_llmobs_spans` with the improved `search_llmobs_spans` tool in the `llm-obs` toolset. - Fixed `include_path` parameter in `get_datadog_trace` (`apm` toolset) to correctly accept a string array. ## March 10, 2026 - Improved `search_datadog_services` performance by switching to a materialized view backend. - Fixed a bug where `analyze_datadog_logs` and other analytics tools could return empty results for longer time ranges. ## March 9, 2026 - Datadog MCP Server is now generally available! - Improved `search_datadog_incidents` with total count, facets, sort options, and advanced query documentation.