rea.run
Compliance
EN 中文

← Environments

Dataset env Public data APIs U.S. public domain HAR

FederalRegisterEnv-REA

FederalRegisterEnv-REA is an AgentEnv replay environment for Federal Register API: U.S. federal rules, proposed rules, notices and presidential documents. Built with REA, published as a Hugging Face dataset, practiced via MCP tools — this page is a teaching mirror on rea.run, not the runtime.

Numbers and fields are cited from the Hugging Face note / catalog snapshot.

Open on Hugging Face

Datasets and large HAR/parquet stay on HF. This page is a teaching mirror only — no runtime here.

Open HF dataset ↗

Recorded target

Product: Federal Register API

Version: v1

License: U.S. public domain

Hosting: public API, real records

Capture: HAR (HTTP Archive)

Updated: Oct 19, 2026 (Asia/Shanghai)

AgentEnv metrics

MCP tools: 10

Retrieval tasks: 10

Downloads (HF snapshot): 236

Likes: 1

id: earakely-scale/FederalRegisterEnv-REA

What this is

FederalRegisterEnv-REA is an AgentEnv replay environment for Federal Register API. Scenario: U.S. federal rules, proposed rules, notices and presidential documents. Teaching mirror on rea.run — practice keywords: AgentEnv, REA, replay environment, MCP tools.

  • Product: Federal Register API
  • Scenario: U.S. federal rules, proposed rules, notices and presidential documents
  • Category: Public data APIs

What REA captured

  • Capture form: HAR (HTTP Archive) (note: Captured as a HAR).
  • Target product / version: Federal Register API v1.
  • License (HF note): U.S. public domain.
  • Hosting / data shape: public API, real records.
  • Large HAR/parquet files live in the Hugging Face dataset — not on these Pages.

What you practice in AgentEnv

This env exposes about 10 MCP tools and 10 retrieval tasks (HF note snapshot). Retrieval tasks: the agent uses MCP tools to find or act on information inside the replay environment; the baseline table shows published model scores. How to read baselines: scores look like a/b — a = tasks correct, b = task count; cross-check with the task total above. Practice runs on AgentEnv / HF, not on rea.run.

Metric Value (HF note)
MCP tools 10
Retrieval tasks 10

Model baselines

Scores cited from the HF note snapshot (retrieval-task scores). How to read: numerator = tasks correct, denominator = task count.

Model Score
gpt-oss-120b 9/10
GLM-5.3-Flash 9/10

Relation to REA

Pipeline: REA captures an authorized browser/API session → published as a Hugging Face dataset → replayed as MCP tools on Scale AgentEnv. rea.run provides the categorized catalog and teaching copy — it is not the runtime and does not host capture files.

  1. 1

    Authorize the target (own / OSS / written permission / approved lab).

  2. 2

    Capture the browser session with REA (HAR or capture-browser-scenario).

  3. 3

    Replay as AgentEnv MCP tools on Scale AgentEnv / the HF space — not on rea.run.

License & compliance

License (cited from HF note): U.S. public domain. Authorized use only: own systems, OSS, written permission, or an approved lab. Do not record or replay against unauthorized production targets. This site does not provide cracking, bypass, or unauthorized-access guidance.

Compliance · Legal boundary

Next steps

Install AgentEnv / REA

To record sessions locally with REA, install rea-agents from official sources. AgentEnv replay and eval run on Scale / HF — not on this site.

HF note (cited)

Snapshot text from the catalog note — not a verbatim dump of a long HF card.

Public data APIs · Federal Register API v1 (U.S. public domain), public API, real records. U.S. federal rules, proposed rules, notices and presidential documents. 10 tools, 10 retrieval tasks; gpt-oss-120b 9/10, GLM-5.3-Flash 9/10. Captured as a HAR.

Same category