webfetch

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/ashlrai-webfetch

About

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list, install config and code. Checked Oct 25, 2026.

Tools

16 read · 1 write · 0 delete · 5 not yet classified

Secrets

5 (API keys, tokens or passwords)

Install scripts

None found

Runs as

Node.js package

Tools (22)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

• analyze_deduplication_quality

  Analyze deduplication quality for a set of ImageCandidates by orchestrating semantic + pHash clustering and surfacing actionable metrics. Returns per-cluster false-positive risk (same hash, different visual — threshold too permissive), false-negative risk (same visual split across clusters — threshold too strict due to compression/alt-source), composite confidence (pHash similarity + metadata similarity), provider diversity, and a recommended threshold. Also generates a flat export-ready metrics table. phashThreshold: maximum Hamming distance (default 8) to merge two candidates. semanticWeight…

• audit_license_consensus

  Reconcile licenses across multiple ImageCandidate objects that may represent the same image from different providers (e.g. wikimedia=CC_BY_SA, unsplash=UNSPLASH_LICENSE). Candidates are grouped automatically by pHash proximity (or URL fallback). For each group the tool: (1) computes pairwise Levenshtein similarity on license strings, (2) determines consensus (similarity > 0.85) or flags a conflict, (3) emits a per-provider audit trail with reasoning sourced from licenseAuditTrail when available, (4) returns a composite confidence score that decays with disagreement and low per-candidate confid…

• batch_cluster_by_phash

  Cluster a set of ImageCandidates by perceptual similarity using Hamming distance on their pHashes. Candidates are grouped using single-linkage clustering (connect when Hamming distance ≤ hammingThreshold). Within each cluster, the best candidate is promoted to 'representative' — ranked by: license openness (CC0 > CC_BY > … > UNKNOWN), then phash confidence, then image resolution, then provider priority (wikimedia > openverse > unsplash > …). Returns: clusters (sorted largest-first then by top score), clusterMetrics (avgClusterSize, lonelyCount, discardedCount), and dedupeRate (fraction of cand…

• batch_find_similar

  Reverse-image-search multiple public image URLs in a single call, across one or more providers (brave, serpapi). Returns per-URL candidate lists with heuristic licenses, deduplicated across providers. Federation-level rate-limit state is checked per-provider per-URL; saturated providers are skipped with a warning rather than blocking the batch. Use when an agent needs to discover similar content for many images at once.

• batch_find_similar_with_distances

  Batch reverse-image search: given multiple reference image URLs, find visually similar images across providers (brave, serpapi) and return results grouped by perceptual-distance clusters. Clusters: 'exact' (Hamming 0-3), 'near-duplicate' (4-8), 'similar' (9-15), 'loosely-related' (16-25). Within each cluster, candidates are ranked license-first (open > platform > editorial > unknown), then by pHash confidence, then provider quality. Returns per-reference pHash metadata, clusters with ranked candidates, aggregate statistics (band breakdown, median/p90 Hamming distances), and non-fatal warnings.…

• compare_candidates

  Analyse a SearchResultBundle for cross-provider duplicates. Runs two detection passes: (1) URL-level — normalised URLs that appear from multiple providers form a 'url' group (confidence 1.0); (2) pHash-level — candidates whose perceptual hashes are within a Hamming distance threshold form 'phash' groups. Returns a structured ProviderDedupeReport: { duplicateGroups: [{members, reason, confidence}], merged }. Use this after search_images or search_artist_images to understand which providers returned the same image, to tune federation provider selection, and to inform cache strategy. hammingThres…

• compare_phashes

  Download two image URLs, compute structured perceptual hashes for each (DCT pHash via sharp when available, aHash fallback otherwise), and return per-image metadata plus the Hamming distance. `isMatch` is true when distance ≤ 6 (near-duplicate). Use to rank candidates by perceptual stability, choose between fast (aHash) vs accurate (DCT) deduplication, and build confidence-aware deduplication pipelines.

• compute_federation_fallback

  Given a FederationRepairPlan (from a prior searchImages call with repairPlan:true) and the detected failure patterns, compute an ordered list of concrete fallback provider substitutions that match the original license policy. Patterns handled: all-unknown-license (switches to providers with structured license metadata, filtered by open-only vs safe-only policy), auth-missing (auto-disables paid/credentialed providers, suggests free alternatives), all-timeout (adds managed-browser opt-in and lower-latency providers), partial-failure (promotes healthy providers, adds backup providers for the fai…

• download_image

  Download an image URL (typically from a prior search_images result) to the local disk cache. Streams with a 20MB hard cap, content-type guard, SHA-256 hash, and returns the cached file path. Host blocklist enforced.

• export_cache_for_replay

  Export cached images as a tar archive suitable for CI test fixtures and local-first development. Optionally filter by mimeType prefix (e.g. 'image/jpeg') or ageMs (only include entries younger than this many ms). The tarball can be imported later with importCache / POST /v1/cache/import to restore a deterministic cache state without network calls.

• extract_image_metadata_audit

  Extract embedded metadata (EXIF artist/copyright, XMP dc:creator/dc:rights/cc:license/xmpRights:UsageTerms, IPTC record 2) from downloaded image bytes, cross-check against provider-supplied candidate metadata, and return a full ImageMetadataAuditResult with per-field confidence scores and a conflict resolution log. Input: base64-encoded image bytes (obtain via download_image) + the ImageCandidate used to find the image. Conflict detection: compares embedded vs provider values using Levenshtein similarity. When similarity > 0.8 (near-agreement), fields are merged and confidence is boosted to 0.…

• fetch_with_license

  Given an arbitrary URL (image or webpage), determine its license via host heuristics + page metadata (<link rel=license>, dc.rights, og tags). Set probe: true to also download the bytes. Use when an agent already has a URL and needs a go/no-go decision before shipping.

• find_similar

  Reverse-image-search: given a public image URL, find visually similar images. Requires SERPAPI_KEY env var and providers: ['serpapi']. Returns candidates with heuristic licenses — treat results as leads, not shippable.

• get_federation_diagnostics

  Returns real-time aggregated provider telemetry over a sliding 5-minute window. Shows per-provider avg latency, result counts, error rates, last-success timestamp, and rate-limit bucket state (saturated / nextTokenAt). Use to detect slow/failing providers, spot rate-limit saturation, and optimize provider selection mid-run. windowMs defaults to 5 minutes (300000).

• get_federation_health_report

  Returns a comprehensive federation health report combining real-time diagnostics with actionable provider recommendations over a sliding 5-minute window. Includes: per-provider health status (healthy/degraded/saturated/unavailable), ranked provider list by composite score (latency_p50, error_rate, throughput), suggested fallbackChain ordered by reliability (excludes saturated/unavailable providers), estimated recovery time from rate-limit buckets, and the full diagnostics dashboard. Use this instead of get_federation_diagnostics when you need both health analysis and routing recommendations in…

Signals

Stars

3

Forks

0

Open issues

1

Last commit (this folder)

Sep 29, 2026 (1 month ago)

Last release

May 10, 2026 (5 months ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name

io.github.ashlrai/webfetch

Version

0.1.5

Auth

API key

Packages

npm getwebfetch-mcp

License

MIT

Repository

github.com/ashlrai/webfetch

Website

getwebfetch.com

First seen

Oct 24, 2026

Last checked

Oct 25, 2026

Source

Official MCP Registrygithub-graphql