Delivery

We provision read access to Google Cloud Storage (GCS) buckets. You authenticate with a service account we provide and pull data on your own schedule.

Delivery Model

Each delivery is a delta. A timestamped folder contains only the signals detected since the previous delivery — not a cumulative snapshot. To build historical coverage, union folders backwards.

Deliveries are immutable. Once a folder lands in a bucket, it's never modified. If data needs to be re-delivered for a past period, it arrives as a new timestamped folder.

Bucket versions (v1, v2, v3) are schema evolutions. The underlying signal data is the same — field names, types, or structure changed between versions. Looking far enough back may require reading from an older bucket version with a different schema. We're standardizing schemas across all signal types in 2026.

Backfill folders are historical deltas. Delivered at onboarding to provide coverage for time periods before your contract started.

Available Buckets

Each signal type has its own dedicated bucket. Access each bucket directly by its URI.

SEC Filings

Social & Web Signals

New Signals (Q1 2026)

Company Intelligence

Reference Data

Bucket Structure

Each bucket contains timestamped folders. Each folder has two files: output.jsonl and output.parquet.

gs://autobound-news-v4/
├── 2026-08-27-09-04-12/
│   ├── output.jsonl
│   └── output.parquet
├── 2026-08-26-09-04-50/
│   ├── output.jsonl
│   └── output.parquet
└── ...

Pull from the most recent folder to get the latest data.

Authentication

1. We provide you with a GCP service account JSON key file
2. Set the environment variable: export GOOGLE_APPLICATION_CREDENTIALS="/path/to/key.json"
3. Use gsutil to access the buckets

# List folders in a bucket (access each bucket directly by URI)
gsutil ls gs://autobound-news-v4/

# Download JSONL
gsutil cp gs://autobound-news-v4/2026-08-27-09-04-12/output.jsonl ./

# Download Parquet
gsutil cp gs://autobound-news-v4/2026-08-27-09-04-12/output.parquet ./

📘

Tip: You can discover all available signal buckets by running gsutil ls -p autobound-signal-delivery. This returns the names of all signal buckets in the project — including buckets you are not licensed for. You can browse bucket names and folder timestamps freely, but reading data inside a bucket requires objectViewer on that specific bucket, which is granted during onboarding. Unlicensed buckets return AccessDeniedException: 403 when you attempt to read their contents.

File Formats

Both formats contain the same data. Choose based on your pipeline.

{"signal_id":"7dfdb4b4-c0b4-4620-aca6-e7263123028e","signal_type":"news","detected_at":"2026-01-15T10:30:00Z","association":"company","company":{"name":"Acme Corp","domain":"acme.com"},"data":{"summary":"Acme Corp announces expansion into European markets...","source_url":"https://example.com/news/acme-expansion"}}

signal_id: STRING
signal_type: STRING
signal_subtype: STRING
detected_at: TIMESTAMP
association: STRING
company: STRUCT<name, domain, linkedin_url, industries, employee_count_low, employee_count_high, description>
contact: STRUCT<first_name, last_name, full_name, email, job_title, linkedin_url, city, state, country>
data: STRING (JSON-encoded)

Refresh Cadence

📘

February 2026 Update: SEC filings, earnings transcripts, news, hiring trends, hiring velocity, and work milestones now deliver weekly (shifted from monthly in February 2026). See the Changelog for full details.

Getting Started

1. Contact [email protected] to get your service account credentials
2. Choose which signal types you need
3. Set up authentication and start pulling data