Skip to main content
Glama

Server Details

Web search, page reading and FLOP miner reliability for agents, priced per call in FLOP.

Ownership verified
Status
Healthy
Uptime
100.0% over the last 24 hours
Last Tested
Transport
Streamable HTTP · MCP 2025-11-25
URL
Repository
0xBobaAI/flop-scout
GitHub Stars
0
ChatGPT
–
Claude
–

TDQS

A3.9/5.0

Scored across 9 tools

Disambiguation4/5

Each tool maps to a distinct job: alert create/read, miner ranking/history, network-level inference cost, catalog discovery, price list, request quoting, and the actual answer call. The main soft spot is the cluster of cost-related tools (scout_pricing, scout_quote, flop_inference_prices), where scout_quote's inference estimate overlaps with what flop_inference_prices returns. Descriptions do enough to steer selection, but an agent must read carefully to pick the right one of the three.

Naming Consistency4/5

Names are uniformly snake_case with stable domain prefixes (flop_ for network/miner data, scout_ for service operations), which makes grouping predictable. The one deviation is grammatical style: some names are verb-led (flop_alert_check, flop_alert_subscribe, scout_answer) while others are noun phrases (flop_miner_ranking, scout_catalog, scout_pricing, scout_quote), so the action is not always legible from the name alone.

Tool Count5/5

Nine tools is well matched to the surface: two alert operations, three miner/network data tools, and four scout service operations covering discovery, pricing, quoting, and execution. Nothing feels redundant or padded; each tool has a clear lane.

Completeness3/5

The core workflow of discover → price → answer is fully covered, and miner data has both ranking and history. However, alerts are create-and-read only: cancelling requires using the REST API by hand, so the alert lifecycle is a dead end inside this server with no update/unsubscribe or list-all-alerts tool. There is also no way to fetch history for several miners at once, forcing repeated calls for comparisons.

Available Tools

9 tools
flop_alert_checkA
Read-onlyIdempotent
Inspect

Return active alerts and their current state. Each item has alert_id, address, threshold (the configured trigger level), settle_rate (current observed rate), and fired (true when settle_rate is below threshold). Poll this after flop_alert_subscribe with channel="site" to detect firings. Use alert_id from the response to cancel an alert via the REST API.

Inputs: address.

flop_alert_subscribeAInspect

Create an alert that fires when a miner's settle_rate drops below threshold. Returns alert_id. Use address "*" to watch every miner at once. channel="site" requires polling with flop_alert_check to detect firing; channel="telegram" pushes a message to handle when the threshold is crossed. threshold is a fraction 0–1 (e.g. 0.5 fires when settle_rate drops below 50%).

Inputs: threshold (required), handle, address, channel.

flop_inference_pricesA
Read-onlyIdempotent
Inspect

Return every live FLOP inference model with the gateway's cost estimate for one request: cost_flop, input_tokens, output_tokens, gflops, latency_blocks. Needs no key and spends nothing. Use to compare miner inference costs before choosing a model for scout_answer, or to show the user what running a FLOP model costs at the network level.

Inputs: max_tokens.

flop_miner_historyA
Read-onlyIdempotent
Inspect

Return the settle_rate time series for one miner from Scout's 15-minute snapshots. Each point has ts (Unix timestamp) and settle_rate (0–1). A short response span means the miner was recently added, not that it is stable. Use window to zoom in on a recent period, or points to control resolution. Useful for trend analysis and alerting context.

Inputs: address (required), points, window.

flop_miner_rankingA
Read-onlyIdempotent
Inspect

Return FLOP inference miners ranked by reliability. Each entry includes address, settle_rate (fraction of on-chain tasks settled, 0–1), local_success (Scout's own observed success rate), avg_latency_ms, and score. Higher score is better. Use to pick a reliable miner, monitor network health, or feed into an alert subscription.

Inputs: no named parameters.

scout_answerA
Read-only
Inspect

Search the chosen sources, synthesise a cited answer, and return an itemised price. The response contains answer (text), citations (list of {url, title, snippet}), charged_flop (Scout fee charged), and inference_estimate (miner cost). FLOP models require a gateway key in keys.flop; OpenRouter models use keys.openrouter. On devnet keys can be omitted — charges are nominal and nothing is debited.

Inputs: q (required), n, keys, model, +2 more.

scout_catalogA
Read-onlyIdempotent
Inspect

List every source (web, scholar, news …), preset bundle, and inference model available. Returns sources (id, label, free flag), presets (id, included sources), and models (model id, provider, context window). Call once to discover valid values for the sources and model parameters of scout_quote and scout_answer.

Inputs: no named parameters.

scout_pricingA
Read-onlyIdempotent
Inspect

Return the full Scout price list: cost per source call, per page read, per token, and the base fee — all in FLOP. Also reports metering_mode ("nominal" on devnet means charges are computed and shown but never debited) and which endpoints are healthy. Call to show the user current rates or to verify the service is operational before a workflow.

Inputs: no named parameters.

scout_quoteA
Read-onlyIdempotent
Inspect

Price a request before running it; spends nothing and debits nothing. Returns total_flop (Scout service fee), and when q and model are given, also inference.floor and inference.typical (the FLOP gateway's own estimate of the miner cost for a minimal and a three-source call respectively). Scout's fee and the miner's cost are always two separate line items. Call this before scout_answer to decide whether to proceed.

Inputs: q, model, sources, max_tokens.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      B
      maintenance
      Lets AI agents read web pages and batches of URLs with prompt-injection screening, search the web, get cited answers, fact-check claims, and check crypto tokens for scam and rug-pull risk. Each call is paid per use in USDC through x402 (from $0.002), with a spend cap.
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI agents to read web pages as clean markdown or JSON, crawl sites, search the web with results already read, monitor pages for changes, and run multi-step browser tasks under configurable limits.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables AI agents to perform reliable web searches through multiple vendors with automatic failover, keyless-first access, TTL caching, and concurrent request coalescing.
      1
      MIT
    Try in Browser

    Glama MCP Gateway

    Add one secure layer between your agents and this server.