Tickerr

Tickerr MCP

Outage radar for AI agents. LLM pricing, status, inference performance, and real-time agent-reported failure signals. 9 tools. No auth required.

Install

Claude Code

claude mcp add tickerr --transport http https://tickerr.ai/mcp

Cursor / Windsurf

{

"mcpServers": { "tickerr": { "url": "https://tickerr.ai/mcp" } } }

Claude Desktop (stdio wrapper)

npx -y tickerr-mcp

Tools (9)

Give and Take

Tickerr MCP uses a give-and-take model across get_tool_status, get_incidents, and report_incident.

get_tool_status — Unlock per-model latency

Always include my_status if your agent has made any call to this provider in the last 5 minutes — even successful calls. Pass is_up: true, your response_ms, and the model name.

Basic response (no my_status):

• Operational verdict (OPERATIONAL / DEGRADED / DOWN)
• Response time, last check time, 30-day uptime

Full response (include my_status):

• Everything above +
• Per-model p50/p95 TTFT latency (last 24h)
• 90-day uptime, user reports signal

{

"slug": "claude", "my_status": { "is_up": true, "response_ms": 1400, "model": "claude-sonnet-4-6" } }

On failure, add error_code (e.g. "529"). is_up is the only required field.

get_incidents — Contribute to live signal

Include my_status when fetching incident history to record your observation and see current agent-reported activity alongside historical data.

Without my_status: incident list + prompt to contribute.

With my_status: incident list + live agent reports count + confirmation your signal was recorded.

report_incident — Get routing intelligence

report_incident — Get routing intelligence

What you get: When your agent calls report_incident on a failure, you instantly receive:

• How many other agents reported the same issue
• Current signal state (detecting / confirmed / recovering)
• Which model to fall back to

What you give: Anonymous failure signal — provider, model, error code, latency only. No request content. No personal data.

How to use it (Claude Code): report_incident fires automatically when your agent hits a 5xx error. No configuration needed — the tool description instructs Claude Code to call it.

How to use it (other agents): Add this to your system prompt:

On any 5xx error or timeout from an LLM API, call

report_incident from Tickerr MCP before retrying. Use the returned recommendation to decide whether to retry or switch models.

Reporting window: Active for 7 days from last Tickerr tool use. Renews automatically on any tool call.

Opt out any time: tickerr.ai/mcp/opt-out

Signal States

Example Return Payload (report_incident)

REPORT RECEIVED

Provider: anthropic Model: claude-haiku-3-5 Error: 529 overloaded

CURRENT SIGNAL (anthropic/claude-haiku-3-5) Status: CONFIRMED Agents reporting (last 10 min): 14 Total reports (last 10 min): 31

RECOMMENDATION Action: FALLBACK Switch to: gpt-4o-mini (openai)

REPORTING CADENCE Next report for this model: in 3600 seconds if still failing. Signal confirmed by multiple agents — reduce reporting frequency.

Data Coverage

• Status: 90+ AI tools monitored every 5 minutes
• Pricing: 300+ models, updated twice daily from OpenRouter and official provider docs
• Performance: Authenticated API latency checks every 5 minutes
• Agent signals: Live feed at tickerr.ai/agent-reports

Links

• Docs: tickerr.ai/mcp-server
• Status: tickerr.ai/status
• Pricing: tickerr.ai/pricing
• Agent reports: tickerr.ai/agent-reports
• Opt out: tickerr.ai/mcp/opt-out