Open Source · MIT License · No Coding Required

Civic Newsroom

An open-source prompt system for turning local public records into structured leads, evidence packs, and draft civic reporting

Since 2005, more than 2,500 U.S. newspapers have closed. City councils still meet. Budgets still get passed. Contracts still get awarded. But in most communities, nobody is reading any of it.

Civic Newsroom is a human-operated workflow for people who want to examine city agendas, budgets, contracts, and public records more systematically. It uses modular AI prompts to help detect story leads, organize supporting evidence, and draft explainers or articles for review.

Human review required. The system helps with discovery, organization, and drafting. It does not independently verify truth or replace editorial judgment.
See How It Works Read the Quickstart View the Prompt Collection View on GitHub
Detect public-record story leads
Cross-check contested claims
Draft civic stories for review
Open-source prompt workflow

How the workflow works

Step 1
Detect
Scan agendas, public documents, and municipal sources for potentially important developments.
Step 2
Verify
Run structured counter-search, source review, and issue checks before trusting a claim.
Step 3
Draft
Produce a clear, human-reviewed writeup, explainer, or story draft.
Show implementation details

Implemented through nine modular prompts: Lead detection · Story expansion · Signal discovery · Adversarial verification · Integrity review · Legal guidance · Plain-language rewrite · Grounding · Standalone research and writing

Public records
→
Detect
→
Verify
→
Draft
→
Human review
→
Share or publish

What you get

Lead list

Prioritized developments worth checking

Evidence pack

Source-linked notes, claims, and contradictions

Draft story

A readable explainer or article draft

Hold report

Reasons a story should not be published yet

Workflow Protections

Civic Newsroom is designed to reduce common AI reporting failures by forcing more explicit sourcing, counter-checking, and review steps. It does not guarantee correctness. It makes weak or unresolved work easier to spot before publication.

Structured counter-checking

Contested or ambiguous claims should be searched across multiple source types with documented query variations before they advance.

Primary-source grounding

Claims should be tied back to original public records whenever possible, not just other coverage or commentary.

Suppression and hold paths

Weak, contested, or incomplete stories should be flagged, held, or suppressed with a visible reason.

Independent review step

Drafts should pass through a separate review prompt or operator check before publication.

Shared source standards

The workflow distinguishes stronger and weaker source categories so the operator can judge confidence more consistently.

Get started

  1. Choose one city or institution to track.
  2. Gather a small set of public source URLs.
  3. Run the lead-detection prompt in a large-context AI model.
  4. Review the strongest leads manually.
  5. Use the verification and grounding prompts on anything contested or important.
  6. Draft a story or explainer only after the evidence is clear.
  7. Review again before sharing or publishing.
Read the Quickstart Guide

Same method, two delivery formats

Civic Newsroom

An open-source prompt collection. Copy and paste into Claude, Gemini, or another large-context AI model. Full control over your data and workflow. Free.

View Civic Newsroom

Civic Transparency Toolkit

A free desktop app that runs the same workflow with automated pipeline execution, source management, and one-click export. Easier to use, no copy-pasting required.

View Civic Transparency Toolkit
Both use the same core workflow and both require human review before publication.

Who this is for

Best for users willing to read source material and review AI output before sharing it publicly.

Engaged Residents

Want to understand what your city council is doing with public funds and how decisions affect your neighborhood.

Transparency Advocates

Believe public records should be genuinely accessible, not just technically available.

Civic Organizations

Need structured analysis of local government decisions to inform community action.

Journalism Students

Learn by working with real municipal data, not textbook examples.

Community Groups

Want to track public decisions that affect your members — zoning, budgets, permits, and policy changes.

Anyone Who Cares

You don't need a journalism degree. You need to care about your town.

What this is not