Request early access

Agent Ops

What's broken. Since when. What changed.

One place for ops engineers and founders whose AI agents already serve customers. Track agents across platforms, see incidents as they start, and follow the change that caused them.

Request early access See how teams would use it →

Public snapshot · as of Nov 12, 12:30 PM CT

Public status pages + rankings tracker

Providers watched

0

Official public status feeds in this snapshot

Incidents, last 7 days

0

2 open now · includes incidents still unresolved

Median time to resolve

0min

Resolved incidents in the last 7 days

Sources: OpenAI (status.openai.com); Anthropic (status.anthropic.com); GitHub / Copilot (githubstatus.com); Pinecone (status.pinecone.io); ElevenLabs (status.elevenlabs.io); xAI / Cursor (status.x.ai); Google Cloud (status.cloud.google.com); Amazon AWS (status.aws.amazon.com); Microsoft Azure (azure.status.microsoft). Times are Central Time.

What's broken. Since when.

Current and recent incidents from official public status pages, last 7 days. 28 in the last 7 days · 2 open.

Public status pages

No incidents in the last 7 days from Google Cloud (Gemini / Vertex AI), Amazon AWS (Bedrock / AI), Microsoft Azure. Meta is not included: no official public status page or API was found.

What changed.

Latest confirmed releases from the Agent Ready rankings tracker, one per tool.

Rankings tracker
  • Nov 12, 2026

    LangGraph 1.2.21

    langgraph 1.2.21 uploaded to PyPI (2026-11-12).

  • Nov 12, 2026

    Anthropic Claude API

    * We’ve launched self-hosted Claude Managed Agents. You run the whole Managed Agents service, not only

  • Nov 11, 2026

    CrewAI 1.15.40

    1.15.40 published on GitHub Releases (2026-11-11).

  • Nov 10, 2026

    GitHub Copilot

    GitHub App support for Enterprise Copilot seat management (2026-11-10).

  • Nov 9, 2026

    Windsurf v3.12.43

    Changed * Devin Local keeps working more often: a response that stops at the output limit before a tool call continues the turn, and GPT models no longer stop to ask for input they can work out. * Image reads in Devin…

  • Nov 9, 2026

    Claude Code v2.1.322

    v2.1.322 published on GitHub Releases (2026-11-09).

  • Nov 9, 2026

    ElevenLabs Scribe v3

    Scribe v3 Scribe v3 and Scribe v3 Live, the most accurate ElevenLabs speech recognition models, are now available in the Speech to Text API.

  • Nov 9, 2026

    LangSmith 2026-11-02 - LangSmith Cloud update

    2026-11-02 - LangSmith Cloud update (2026-11-09).

  • Nov 8, 2026

    Pinecone Search by meaning in pg_pinecone

    Search by meaning in pg_pinecone pg_pinecone can now embed a table's text column with a hosted model, such as

  • Nov 6, 2026

    Cursor Voice Chat with Agents

    Voice Chat with Agents (2026-11-06).

  • Nov 5, 2026

    OpenAI API

    Feature · Model: gpt-6-cyber Use GPT-6-Cyber to investigate vulnerabilities, reproduce them in an authorized test environment, and verify fixes. See

  • Nov 2, 2026

    OpenAI Realtime API

    Feature · Model: gpt-live-1 · API: v1/live/sessions Outbound SIP calling is now available to all organizations through the Live API. Connect GPT-Live to a phone number through your SIP provider. See the

  • Nov 2, 2026

    xAI Grok API

    grok-imagine-image-quality retirement on November 2 On November 2, 2026, `grok-imagine-image-quality` is retired. Requests to the slug will be served by `grok-imagine-image-2.0` with `quality` set to `low`, with no ch…

  • Oct 29, 2026

    Arize Phoenix 21.0.0

    Arize Phoenix 21.0.0 major release (GitHub, published 2026-10-29 10:29 PM CT); client 3.6.0 shipped alongside.

Source: local rankings tracker at agent-ready-rankings (last published to rankings.leffke.com). Links go to each tool's release notes.

Cost per run, last 7 days

Watch spend drift, then see what changed when a run gets expensive.

Example data — illustrative

Nov 6–12, 2026 · example series

Current (Nov 12)

$0.33
↓ from $0.92 midweek after a retry loop was caught
Example cost per run over seven days Illustrative costs: Nov 6 $0.41, Nov 7 $0.43, Nov 8 $0.44, Nov 9 $0.92, Nov 10 $0.87, Nov 11 $0.36, Nov 12 $0.33. $1.00 $0.70 $0.40 Retry loop Nov 6 Nov 7 Nov 8 Nov 9 Nov 10 Nov 11 Nov 12 6 7 8 9 10 11 12

How to read it: midweek spend jumped when an order-status agent retried a failed tool call in a tight loop. After the loop was stopped, cost per run fell back to $0.33. This series is illustrative, not from a live account.

Before and after

Illustrative scenarios — not customer quotes, and not named companies.

Example data — illustrative

Scenario A · Support-desk agent

Time to detect a failure

40 min→4 min

A customer-facing support agent started failing silently after a model change. Before, the team noticed when ticket volume spiked. In this scenario, the failure appears in the incident feed in four minutes — with the model change already attached as the likely cause.

Scenario B · Voice receptionist

Time to recover

2 hr 10 min→18 min

A voice receptionist hung on a downstream tool timeout. In this scenario, the change log points to a deploy eleven minutes earlier, so the rollback starts immediately instead of after a long hunt through logs.

Scenario C · Order-status agent

Cost per run after a retry loop

$0.91→$0.33

An order-status agent retried a failed tool call overnight. In this scenario, the cost-per-run sparkline flags the spike the next morning, the loop is stopped, and spend returns to its usual range.

Request early access

If you run production agents that serve customers and want one place for what's broken, since when, and what changed, tell us how you work today.

Email rick@leffke.com