Feeds/Feed
Live Feed Connectivity

Unlisted feed: shared by direct link only. Not shown in the public marketplace directory.

AI & Machine Learning Papers (arXiv)

ID:da6b2e9a-342a-410c-9d40-399bff5898a1
Alias:
ai-papers-v3
Alex  Lin
Alex Lin
Provider
Feed Type
Collection
Standard RESTful
Verified Source

Curated, machine-readable feed of recent computer science and machine learning preprints from arXiv. Records are normalized to a stable JSON schema suitable for agents and analytics pipelines, including metadata, canonical links, abstracts, and a concise one-sentence insight. 1. Data is ingested on a scheduled basis via the provider push API; 2. Consumers retrieve the current snapshot through the standard read API with cursor-based pagination.

Consumption Rate
$0.008/ 1k calls
Request Custom Feed

SECURED VIA STRIPE CONNECT

Cost Predictor
Calls/Day:
Estimated Cost:
$0.24/ mo

Data Health & History

30D Reliability
52%
Live Heatmap
Last Sync
4/5/2026
Delayed 8202 min
Sync Interval
Daily
Guaranteed refresh rate

Agent-Native Workbench

Connected

HTTP REST Test Console

Use a real API key and query parameters to hit this feed.

Sent as the Authorization header.Add Payment Method

Not Callable - Insufficient credits for one call

This metered feed can be called with enough credits. Add a payment method only for uninterrupted access after credits are exhausted.

Appended to the URL as ?limit=.Max 100

Sent as ?jsonpath= for server-side JSON filtering.Learn Syntax

Add this skill to your OpenClaw skill directory to enable autonomous data fetching.

View Docs

One-line install

curl -fsSL https://pipeagent.dev/install.sh | bash -s -- --alias "ai-papers-v3"
1---
2name: "ai-papers-v3"
3description: "PipeAgent: Curated, machine-readable feed of recent computer science and machine learning preprints from arXiv. Records are normalized to a stable JSON schema suitable for agents and analytics pipelines, including metadata, canonical links, abstracts, and a concise one-sentence insight.
4
51. Data is ingested on a scheduled basis via the provider push API;
62. Consumers retrieve the current snapshot through the standard read API with cursor-based pagination."
7license: "MIT"
8emoji: "⚡"
9user-invocable: true
10requires:
11 env:
12 - PROVIDER_READ_API_KEY
13 config:
14 - feed_id: "da6b2e9a-342a-410c-9d40-399bff5898a1"
15 - base_url: "https://api.pipeagent.dev"
16tools:
17 - type: function
18 function:
19 name: "fetch_pipeagent_da6b2e9a"
20 description: "Fetch live structured data from PipeAgent for feed da6b2e9a-342a-410c-9d40-399bff5898a1."
21 parameters:
22 type: object
23 properties:
24 jsonpath:
25 type: string
26 description: "Optional JSONPath to filter/project results."
27 required: []
28 # The HTTP call this tool should execute
29 x-pipeagent-http:
30 method: GET
31 url: "https://api.pipeagent.dev/v1/feed/da6b2e9a-342a-410c-9d40-399bff5898a1"
32 headers:
33 Authorization: "Bearer {{PROVIDER_READ_API_KEY}}"
34---
35
36### Instruction
37You are an expert data analysis agent. Use this skill to fetch live structured data from PipeAgent.
38The endpoint results are pre-parsed JSON.
39
401. Use the provided tool definition to call `GET https://api.pipeagent.dev/v1/feed/da6b2e9a-342a-410c-9d40-399bff5898a1` with the configured `feed_id`.
412. If the user provided a `jsonpath`, pass it as the `jsonpath` argument to minimize context window usage.
423. Present the findings in a clean, professional table format.
434. Validate and reason about the response according to the JSON Schema below.
44
45**[SILENT EXECUTION]**
46- Do not explain your technical steps.
47- Do not describe the API call process.
48- Only return the final, user-facing answer.
49
50### Response JSON Schema
51response_schema: |
52 {
53 "type": "array",
54 "items": {
55 "type": "object",
56 "required": [
57 "id",
58 "title",
59 "link"
60 ],
61 "properties": {
62 "id": {
63 "type": "string"
64 },
65 "link": {
66 "type": "string"
67 },
68 "title": {
69 "type": "string"
70 },
71 "authors": {
72 "type": "array",
73 "items": {
74 "type": "string"
75 }
76 },
77 "abstract": {
78 "type": "string"
79 },
80 "captured_at": {
81 "type": "string"
82 },
83 "publish_date": {
84 "type": "string"
85 },
86 "one_sentence_insight": {
87 "type": "string"
88 }
89 }
90 }
91 }
92
Active on 42k Agents

Documentation

Overview

This data product delivers recent arXiv preprints relevant to artificial intelligence and machine learning, presented as a versioned collection (snapshot semantics per sync). Each record is a single paper with stable identifiers, bibliographic fields, abstract text, and a one-sentence insight intended for quick scanning by humans and downstream agents.

Data source and provenance

  • Primary source: arXiv public APIs / Atom export, in accordance with arXiv’s API terms of use.
  • Normalization: Provider-side jobs map Atom entries into the JSON schema declared for this feed. Field names and types are stable across releases unless a schema version bump is published.
  • Insight field: By default, one_sentence_insight is produced with a local heuristic. Optionally, the provider job may be configured to use an external language model; that mode is operational only and does not change the consumer contract.
  • Update cadence

  • Scheduled refresh: The provider crawler runs on a daily schedule (UTC-aligned window as configured in deployment).
  • Collection semantics: Each successful sync corresponds to a new batch in storage; consumers always read the current batch unless historical access is explicitly offered elsewhere.
  • Consumer API

    Authenticated clients retrieve data with:

    http
    GET https://api.pipeagent.dev/v1/feed/{FEED_ID}
    x-api-key: YOUR_READ_KEY

    Bearer authentication is also supported where configured for your account.

    Common query parameters

    ParameterPurpose
    limitPage size (capped by platform policy; default if omitted).
    cursorOpaque cursor for the next page when metadata.has_more is true.
    jsonpathOptional projection of the JSON payload (subject to platform limits).
    idsComma-separated list of business keys (id field) for targeted retrieval when supported.

    Pagination follows cursor-first semantics: rely on has_more and next_cursor rather than inferring totals from partial pages.

    Response schema (records)

    Each element of the top-level array is an object with at least:

    FieldTypeDescription
    idstringStable paper identifier in this feed (e.g. derived from arXiv id / version).
    titlestringPaper title.
    authorsstring[]Author list in display order.
    publish_datestringPublication or announcement date (ISO-8601 date or datetime string).
    abstractstringFull abstract text where available.
    one_sentence_insightstringShort, non-authoritative summary for triage.
    linkstringCanonical URL to the paper on arXiv.
    captured_atstringUTC timestamp when the record was captured for this feed.

    Required fields for validation: id, title, link.

    Example record

    json
    {
      "id": "arxiv:2401.00001v1",
      "title": "Example: On scalable training of transformer models",
      "authors": ["A. Researcher", "B. Scientist"],
      "publish_date": "2024-01-15",
      "abstract": "We study methods for efficient training under resource constraints…",
      "one_sentence_insight": "Proposes practical techniques to reduce training cost while preserving accuracy.",
      "link": "https://arxiv.org/abs/2401.00001",
      "captured_at": "2026-03-23T12:00:00.000Z"
    }

    Provider operations (reference)

    Ingestion is performed by the repository job SelfMadeFeeds/arxiv-papers/crawler.ts, which posts to:

    code
    POST /api/v1/provider/feed/{FEED_ID}/push

    using a write API key. Large runs may split payloads across multiple posts sharing a batch id; refer to the provider integration guide for batch_complete semantics.

    Pricing

    Per-call pricing follows the price per 1,000 calls (price_per_1k) shown on the feed listing. Free promotional credits, if available on your account, are applied before metered charges according to platform billing rules.

    Disclaimer

    arXiv and paper metadata are third-party sources. This feed does not constitute peer review, investment, or legal advice. Verify critical facts against primary publications and official arXiv pages.

    Sync History

    Live Heartbeat
    100.0% Uptime
    24h AgoNow
    Latest Updates
    No recent syncs recorded.