# July 2026 Maple in Slack, alerts rebuilt on one engine, a documented v2 API with an Alchemy provider, PlanetScale, and spend limits. Released July 2026. **Contains breaking changes.** ## Highlights - Maple in Slack — alert delivery, plus an agent you can ask about a live incident - Alerts rebuilt on one engine, with a preview chart that matches the evaluator - A documented v2 API, an Alchemy provider, and browser-based CLI login - PlanetScale connects in one step and shows up in the service map - Spend-first billing, with hard spend limits - Issues, investigations, and escalations in one workflow Maple is in Slack now, alerts were rebuilt on one engine, and there's a documented public API you can drive from code. ## Maple in Slack ![An alert delivered into a Slack #incidents channel, carrying the rule, the triggering value and a link into Maple — followed by someone asking the Maple bot what changed, and the bot answering from real telemetry.](/changelog/2026-07-slack.webp) Install Maple into your Slack workspace from **Integrations** and you get two things. Route any alert rule to a channel. Deliveries carry the rule, the triggering value, and a link into Maple. Mention the bot to ask about services, errors, traces, or a live incident. It answers from your telemetry and posts its status as it works. Reconnecting to refresh permissions no longer means uninstalling first. ## Alerts, rebuilt ![Alerts, rebuilt — one engine, and a preview chart produced by the evaluator that actually runs the rule.](/changelog/2026-07-alerts.webp) **Breaking.** Alerts now run on one engine. On upgrade: - Metric rules became query-builder drafts. Review them before re-enabling. - Rules without a destination were disabled. What's new: - The preview chart now comes from the evaluator that runs the rule — same windows, filters and reducers — so it matches the rule's detail page. Raw SQL rules get one too. - Failed evaluations record an error and write a check row, so gaps surface instead of going silent. Skipped checks, missing destinations and delivery failures all land in one place. - Alerts opens on an overview: health, active incidents, activity. - Email destinations, without wiring up a webhook. - Anomaly detection is less trigger-happy, and traffic-outage false positives are down. ## A public API ![A public API — a documented, stability-committed v2, and an Alchemy provider so your dashboards and rules deploy with your infrastructure.](/changelog/2026-07-api.webp) The **v2 API** is documented and stability-committed, with prefixed object IDs, cursor pagination, a consistent error envelope, and snake_case throughout. It covers API keys, dashboards, alert rules, destinations, incidents, error issues, telemetry queries, ingest keys, attribute mappings, scrape targets, and recommendations. Keys get scopes, expiry, and per-key rate limits, and there's an **API Reference** tab in Settings. On top of that: - **`@maple-dev/alchemy`** — declare dashboards, alert rules, destinations and API keys in your `alchemy.run.ts` and deploy them with the rest of your infrastructure. - **`maple auth login`** — browser-based CLI login, credentials in your OS keychain. - **MCP OAuth** — connect Maple's MCP server from any standards-compliant client. Standard API keys work against MCP, members can create MCP keys, and setup is one click from **Connect**. ## PlanetScale ![PlanetScale — connect with a service token and branch metrics collect themselves, with no scrape target to configure by hand.](/changelog/2026-07-planetscale.webp) Connect your org with a service token and Maple collects branch metrics automatically — no scrape target to configure. Databases and branches show up as inventory, and the service map now draws the PlanetScale database behind Hyperdrive instead of stopping at the proxy. ## Spend limits ![Spend limits — billing you can actually read: what you're spending, what's driving it, and a cap it will not cross.](/changelog/2026-07-billing.webp) Billing is spend-first: what you're spending, what's driving it, and where it's heading, with an estimated cost breakdown by signal and in-app invoices. Set a **spend limit** and Maple stops accepting data past it rather than surprising you at the end of the month. Session recordings are metered and size-capped. ## Investigations ![Investigations — issue to thread to evidence-backed diagnosis to approved action, one workflow that can read your source code.](/changelog/2026-07-investigations.webp) Issues, investigations and escalations are one workflow now: issue → thread → evidence-backed diagnosis → approved action → escalation audit. Investigations can read your source code, so findings point at real lines. The workspace renders structured tool output instead of raw JSON and survives a reload. Escalation policies run on one shared engine, with a simulator in settings. ## Explore The four signals — traces, logs, metrics and replays — now live together under **Explore**. - Every service keeps the same color across charts, traces and the service map. - Trace timeline rework: zoom is clamped to the viewport and hover is much faster. - Duration filtering is anchored in your real distribution, with p50/p95 presets. - Latency percentiles are colored by magnitude. Syntax highlighting in the advanced filter editor. - Traces works on every screen size. - Replays got a player-first layout with a lane-aligned session list. - Service map: database nodes split by identity, Hyperdrive collapses, an environment switcher, Cloudflare zones and Workers as nodes, and no layout shift while it loads. - The metrics explorer is a browse grid with shareable metric detail pages, and attribute suggestions are scoped to the selected metric. ## Dashboards - Templates split into *Ready for your data* and *Needs setup*, and the ones that need setup say what's missing (`no k8s.pod.*`) instead of dimming. - Chart types are unified between raw-SQL and query-builder widgets. - Better default heights, a real loading screen, and a rebuilt list view. - Fixed: stacked areas outlining the whole stack; the "Update failed" banner latching on. ## Setup and integrations - **Setup audit** now runs 41 checks across ingestion, attributes and instrumentation. - **Cloudflare connects over OAuth** and provisions its datasets for you, with an infrastructure page and zone analytics filterable by path, country and client. - Ingestion settings rebuilt around one endpoint-and-keys card. API keys get masked reveal, and scopes and expiry at creation. - Integrations list, empty states and drill-ins redesigned. Kubernetes and host filters run instantly. - Local mode gained metrics, services and errors views plus telemetry archives. - A global chat slide-over replaced the sidebar entry — open it anywhere with `C`. ## Fixes - Fixed the main-thread freezes and tab crashes on data-heavy pages, and added guardrails. - Rerender cascades removed; skeleton and page transitions moved off the main thread. - Trace and dashboard loads are faster, and wide time ranges hit new search indexes. - Transient "Cannot reach Maple API" errors retry themselves. - Hard-refreshing a deep link no longer bounces you home. - Silent data-fetch failures now surface as a retryable error state; crashes show a branded screen. - Fixed duplicate notification emails, broken session playback, and timezone jumps on custom ranges. ## SDKs Session replay is built into the Effect SDK — you no longer need the separate browser package to record sessions (`@maple-dev/effect-sdk` 0.6.0, `@maple-dev/browser` 0.2.0).