Tool catalog

agent_research()

Autonomous multi-step research on a topic — search, read, synthesize.

POST /v1/agent/research

batch_extract()

Extract up to 50 URLs in parallel — async batch job.

POST /v1/batch/extract

bias_detect()

Detect framing bias in news or marketing text.

POST /v1/bias/detect

cache_url()

Check if a URL is cached — 0 units before a full ingest or extract.

GET /v1/cache/url

capabilities()

Platform capabilities for agents

GET /v1/capabilities

chunk()

Split long text into token-aware chunks for RAG.

POST /v1/chunk

compare()

Semantic diff — how much two texts overlap or diverge.

POST /v1/compare

confidence_estimate()

Estimate model confidence on a classification or extraction.

POST /v1/confidence/estimate

conflict_detect()

Find contradictions between two or more text sources.

POST /v1/conflict/detect

context_compress()

Compress conversation history while preserving key facts.

POST /v1/context/compress

crawl()

Async site crawl — returns job ID, webhook on completion.

POST /v1/crawl

crypto()

Secp256k1 ops for agent wallets — sign, verify, address derive.

POST /v1/crypto/invoke

docx()

Extract text from Word .docx files.

POST /v1/docx/extract/upload

embed()

GPU embeddings for RAG — private inference, no third-party LLM.

POST /v1/embed

estimate()

Estimate units, time, and cost for a planned tool chain.

POST /v1/estimate

extract()

Extract clean text or markdown from any URL — core web tool.

POST /v1/extract

extract_json()

Extract JSON blobs embedded in HTML or API responses.

POST /v1/extract/json

extract_links()

All outbound links from a page — crawl planning, sitemap discovery.

POST /v1/extract/links

extract_query()

Run a CSS/XPath-style query against fetched HTML.

POST /v1/extract/query

extract_schema()

Pull schema.org / OpenGraph fields into typed JSON.

POST /v1/extract/schema

fallback_chain()

Try tools in order — pay only for the one that succeeds.

POST /v1/fallback/chain

html_to_markdown()

Convert raw HTML string to clean markdown.

POST /v1/html/to-markdown

image_describe()

Describe image content for multimodal agent pipelines.

POST /v1/image/describe

ingest()

Full RAG ingest: fetch → markdown → chunk → optional embed.

POST /v1/ingest

interact()

Browser interaction — click, fill forms, multi-step page flows.

POST /v1/interact

jobs()

Unified async job status (batch/crawl)

GET /v1/jobs/{id}

json_ld()

Extract JSON-LD structured data from a page or HTML blob.

POST /v1/json-ld

map()

Site map — discover URLs from a root domain.

POST /v1/map

memory()

Store and recall short agent memory snippets by key.

POST /v1/memory/store

metadata()

Open Graph, title, favicon — page metadata without full extract.

POST /v1/metadata

monitor()

Monitor URL for content changes — webhook alert when diff detected.

POST /v1/monitor

ocr()

OCR scanned images and PDF pages — text + confidence scores.

POST /v1/ocr

ocr_table()

Table OCR — structured rows/columns from spreadsheet scans.

POST /v1/ocr/table

pdf()

Extract text and structure from PDF documents.

POST /v1/pdf/extract

plan()

Build an execution plan with budget caps for a multi-step agent task.

POST /v1/plan

readability()

Mozilla Readability — main article content only, no nav clutter.

POST /v1/readability

rerank()

Rerank document list by relevance to a query.

POST /v1/rerank

retrieve()

Semantic retrieval over an in-session document index.

POST /v1/retrieve

screenshot()

Full-page or viewport PNG screenshot via headless browser.

POST /v1/screenshot

search()

Web search with ranked snippets — agent research starting point.

POST /v1/search

search_read()

Search + read top results in one call — deep research shortcut.

POST /v1/search/read

self_reflection()

Agent self-check — does the answer address the original goal?

POST /v1/self/reflection

session_budget()

Track remaining units and spend for an agent session.

GET /v1/session/{id}/budget

similarity()

Cosine similarity between texts — reranking and dedup.

POST /v1/similarity

source_credibility()

Score domain/source trust for fact-checking pipelines.

POST /v1/source/credibility

summarize()

Abstractive summary of long text — agent context compression.

POST /v1/summarize

trace_export()

Export a full trace of tool calls for debugging or audit.

GET /v1/trace/{session_id}

transcribe()

Speech-to-text from audio URL or upload (Whisper GPU).

POST /v1/transcribe

transcribe_diarize()

Transcribe with speaker diarization — who said what.

POST /v1/transcribe/diarize

translate()

Translate text between languages on private GPU.

POST /v1/translate

validate_email()

Syntax + MX check — is this email deliverable?

POST /v1/validate/email

validate_phone()

Parse and validate phone numbers (E.164).

POST /v1/validate/phone

validate_url()

Validate URL format, DNS, and reachability.

POST /v1/validate/url

verify_fact()

Fact-check a claim against web evidence — verdict + sources.

POST /v1/verify/fact

webhook_send()

POST results to your webhook when async jobs complete.

POST /v1/webhook/send

workflow_run()

Run a fixed multi-tool pipeline in one request (billed as child sum ×0.8).

POST /v1/workflow/run

workflow_stateful()

Stateful workflow with variables passed between steps.

POST /v1/workflow/stateful

youtube_transcript()

YouTube video transcript — captions without browser automation.

POST /v1/youtube/transcript