Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

Zarqa logo

Zarqa

CI License: Apache-2.0 Python 3.10–3.12 PyPI Release

An open-source, AI-powered threat intelligence tool. Zarqa collects intel from the internet and from the threat intel platforms you already use, turns it into structured, sourced intel, ranks it by how much it matters to your organization, and answers questions about it with cited evidence.

It is not a replacement for anything. It works next to your existing platforms, reads from them, and can send what it finds back to them in standard formats.

Named after Zarqa al-Yamama, who in Arabian legend could see an army coming days before it arrived.

Beta, solo-maintained. Collection, extraction, enrichment, relevance ranking, investigations, brand watch, sharing, reports and the connector layer work and are tested. First benchmark: benchmarks (without a model it is precise but misses most names and techniques; a small local model adds little). See the roadmap.

What works today

  • Collect from the internet and from other platforms. RSS/Atom feeds (with the full article), pages, PDFs, indicator lists (text, CSV, JSON, STIX), and pulls from other intel platforms through MCP connectors or the built-in STIX/TAXII 2.1 and REST bridges. On a schedule, or by hand. (collection)
  • Turn text into structured intel you can check. Indicators (defanged or not), CVEs, ATT&CK techniques, threat actors, malware, tools, campaigns and targets, and the relations between them. Each one links to the sentence it came from and says how it was found. (extraction)
  • Models are checked, not trusted. The extractor model must quote the document for every item; items whose quote is not in the text are dropped, and a model can label indicators but never add one. Rules and ATT&CK names work with no model at all.
  • Aliases resolved. "Fancy Bear", "Sednit" and "Forest Blizzard" are one actor (APT28), from MITRE ATT&CK data shipped with Zarqa.
  • Ranked for you. Describe your organization and your intelligence requirements; every report gets a 0-100 score with the reasons (your products named, your sector targeted, known-exploited CVEs, a requirement answered). (relevance)
  • Ask, and get a cited answer. "What do we know about this IP?" The analyst model searches the store and your connectors; every finding cites its evidence, unsupported claims are set aside, and "not enough data" is an allowed answer. (investigations)
  • Brand watch. Lookalikes of your domains, resolved and checked: how new they are, whether the page uses your name or asks for a password, or is parked. Scored and tracked with a status. (brand)
  • Enrichment on request. Connectors declare their lookups; built-in EPSS and RDAP are opt-in and never sent your own or internal values. (enrichment)
  • Shares in standard formats. STIX 2.1 and MISP-format exports, a read-only TAXII 2.1 server, pushes to other platforms that a person starts, and an MCP server so other agents can ask Zarqa. (sharing)
  • A brief. What came in over a day or a week, most relevant first, with new names, indicators, known-exploited CVEs, lookalikes and investigations; Markdown or HTML. zarqa brief.
  • Connectors, read-only by default. Writes to another platform go only through a sharing connector with the exact tool allowed, never from a model, always audited. (connectors)
  • Any model. Local models through Ollama, or cloud providers, per role. Prompts to cloud models are masked (emails, private IPs, user names and your own terms).

Screenshots

From the offline demo (fictional data).

A report: what was found, with the sentence each came from An investigation: every finding cites its evidence
document investigation
An actor: where it was seen, relations, what ATT&CK knows Brand watch: lookalikes scored and tracked
entity brand

Planned

A benchmark with a stronger model; registrant comparison in the brand watch; separate worker processes.

Try it

Offline, with fictional sample data and no model: zarqa demo --run, then ZARQA_HOME=~/.zarqa-demo zarqa serve (demo). Or docker compose --profile demo up demo.

pipx install zarqa               # or: pip install zarqa   (extras: zarqa[all])
zarqa init                       # config.yaml, .env (mode 600) and data/ in the home folder
zarqa doctor                     # checks the setup and says what to fix
zarqa serve                      # http://127.0.0.1:8621

zarqa add --url https://...      # a report, extracted straight away
zarqa sources pull               # pull every source in config.yaml now
zarqa lookup 203.0.113.66        # everything known about a value, with the sentences
zarqa ask "What do we know about 203.0.113.66?"

Docker: docker compose up -d (non-root, read-only container, loopback port).

Configuration

config.example.yaml documents every setting. Secrets go in .env only. Login modes: none (loopback only), basic, local, OIDC and reverse-proxy headers, with viewer, operator and admin roles. The server refuses to listen on a network address without a login.

Project

Metadata

Release files for zarqa 0.1.0b0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for zarqa 0.1.0b0
File Size Uploaded Provenance Actions
zarqa-0.1.0b0.tar.gz 278.5 kB With provenance Details for zarqa-0.1.0b0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for zarqa 0.1.0b0
File Interpreter ABI Platform Actions
zarqa-0.1.0b0-py3-none-any.whl Python 3 none any Details for zarqa-0.1.0b0-py3-none-any.whl

Total release size: 554.8 kB

Release files / zarqa-0.1.0b0.tar.gz

Download URL zarqa-0.1.0b0.tar.gz
Size 278.5 kB
Tags Source
SHA-256 checksum
How to use checksums
22462671d17e4517439291d265675c3bf79b2bd90557c3d4c02d602e07d68f09
BLAKE2b-256 checksum
How to use checksums
f18b37ef6088161ae05a0fccf7a91270a5ae39976f314933ec9669db51bd1304
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 17, 2026.

Transparency log

Release files / zarqa-0.1.0b0-py3-none-any.whl

Download URL zarqa-0.1.0b0-py3-none-any.whl
Size 276.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
152716f114073671ad45bf48873b7c90b87374efed60a6cd7570806f45188e3e
BLAKE2b-256 checksum
How to use checksums
e4a50a8144595b2b4cab83911590e08f4fc4f3e4ccc87a27ec64a12b8f71399c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 17, 2026.

Transparency log

Security for zarqa 0.1.0b0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for zarqa-0.1.0b0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 17, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page