For newsrooms & content teams

The editorial workspace, grounded in your own materials.

Narratic is where newsrooms collect sources, track stories, and work with AI — without the AI making things up.

Closed beta EU-hosted Built in a real newsroom
The Narratic editor's desk: pending reviews, unassigned topics, active threads, and a sidebar of sources, topics and tasks.
№ 01 The problem

Three things on every editor's desk before lunch.

The bottlenecks aren't dramatic — they're daily. They're the friction that turns a good story into "we'll get to it next week."

Materials in eight places

Links in Slack, PDFs on Drive, notes on a phone, calendar in Google, transcripts in Word. Come back to the story three weeks later — you rebuild the context from scratch.

Chatbots hallucinate

Generic AI answers from memory instead of from your materials. Every output needs fact-checking before it's usable. In practice — no time saved.

The workflow lives in the editor's head

Who's writing what, what's in review, what ships tomorrow — tracked in group chats, not in a tool. New person joins and nobody can hand it over.

№ 02 The mechanics

Four ideas that change how a team works on a story.

Together they make Narratic feel different from a generic notebook or a chatbot bolted to a CMS.

01

Story-Centric Workspace

Everything orbits the Topic. Sources, comments, calendar, content — one view per story, no context-hunting.

02

Source-Grounded AI

The AI only sees the materials you attach. No internet crawl, no model hallucination, no guessing.

03

Human-in-the-Loop by design

AI suggests. You decide. Every AI action is reviewable, reversible, and auditable — with cost tracking per workspace.

04

Offline-first field work

Reporters work without connectivity. Edits queue locally, then sync the moment the network comes back.

№ 03 How it looks

Four claims, with the receipts.

The mechanics above translate into shipped, daily-used UI. Below — what they look like in a real Polish newsroom.

Topic view: title, status pills, deadline, source list and an AI assistant proposing 10 article ideas grounded in three attached materials.
01 · Story-Centric Workspace

Every story has its own room.

Topic title, status, deadline, the materials it was built from, and an AI assistant that only sees what's attached. Reporters work end-to-end without switching tools or rebuilding context.

Full-screen AI chat with three attachment chips, a user prompt, and a structured response listing article angles grounded in the picked materials.
02 · Source-Grounded AI

AI that points back to the page.

Pick the materials, ask the question. Every answer rests on the attachments you chose — no open-internet crawl, no cross-workspace bleed, every per-call cost logged for audit.

Source detail view: status pills, created and updated dates, an AI-generated one-paragraph summary, and a list of attached files and links.
03 · Materials & auto-summary

A summary on every source.

Each attachment — PDF, link, email, transcript — gets its own one-paragraph TL;DR as it lands, and the source itself rolls up a summary across all of them. Click through to the original whenever you need the actual quote.

Workspace-wide search: a query for 'dziki' returns 11 hits across topics, notebooks, leads and article bodies, with the term highlighted inline.
04 · Find anything

One search across everything.

Topics, notes, attachments, AI summaries, source content — searched by keyword and by meaning. Find a quote from three months ago even when you only remember what it was about.

№ 04 What you get

A foundation that's useful — even before AI.

Ten concrete pieces, split into the editorial basics and the layer that makes them sing.

Foundation

The editorial basics — capture, organize, ship. Useful before you turn on a single AI feature.

Materials panel

Eight source types — PDF, DOCX, XLSX, links, iCal, audio metadata. Automatic URL fetching, full-text search, metadata extraction.

Notebooks

Per-topic research workspace. Write notes, link sources, sketch the story before committing to a draft.

Editorial workflow

Eight-state status machine on every topic, with editor-in-chief approval gate before publish.

Tasks

Assignment, due date, status. Who's writing what, what's in review, what ships tomorrow — tracked in the tool.

Content editor

Draft articles with source citations. Draft → review → published states; cleanly versioned along the way.

Calendar feeds

Subscribe to iCal feeds — council meetings, court dockets, government events. Linked to topics.

AI & teamwork

The layer on top — what makes the foundation hum with a team and an AI assistant.

AI chat with attachments

OpenAI + Anthropic. Streaming responses scoped to the materials you pick. Per-workspace cost tracking.

Threaded comments

Comments on topics, sources, and content fragments. @mentions, resolved state, no thread sprawl.

Permissions

15 role presets × ~50 granular permissions, ALL/OWN scope. A reporter doesn't see the editor-in-chief's materials.

Offline-first SPA

IndexedDB cache, mutation outbox, service worker. Works in the field without wifi; syncs on reconnect.

№ 05 For specific teams

Opinionated by design.

Narratic is not for everyone. We say plainly which teams we serve well — and which we don't.

.

Local newsrooms

Regional dailies, city portals. Small teams, big material flow, real deadlines. One editor manages 20–30 topics in parallel without losing context.

.

B2B content teams

Content marketing agencies and in-house teams. Research → brief → draft → review → publish. Per-client AI cost audit.

.

Research departments

Non-profits, think tanks, analytical teams. Long projects, dozens of sources (reports, hearings, interviews). Source-grounded AI replaces "internet search."

Not for solo bloggers or SEO content farms. Narratic earns its keep when there's a team and a real material flow.

№ 06 Trust

Built for sensitive material.

Newsrooms handle leaks, whistleblowers, and sources who trust them. Narratic is built around that trust.

Workspace isolation

Every team has its own space. No information leaks between organisations.

Granular permissions

15 roles × ~50 permissions, scope ALL/OWN. A reporter doesn't see the editor-in-chief's materials.

AI no-train policy

Materials sent to OpenAI/Anthropic never train models. Per-workspace cost transparency.

Auditable AI answers

Every AI response is grounded in specific materials. Easy fact-check before publication.

EU hosting

Frankfurt data centre (Digital Ocean). GDPR-friendly by default.

Smaller attack surface

Offline-first means less data in transit. Field work is local-first by design.

More detail: Privacy policy · Security note

№ 07 Coverage

Two languages today. The rest of Europe is queued.

Narratic ships in Polish and English. We turn on the next language when a real newsroom is ready to use it — not before.

Live today

  • Polish
  • English

Next in line

  • German Soon
  • French Soon
  • Spanish Soon
  • Ukrainian Soon

Want yours next? Tell us at hello@narratic.app and we'll line it up against the rest.

№ 08 Closed beta

Built in a real newsroom. Now open to yours.

We're onboarding editorial teams one by one. Tell us who you are and what you're working on — we'll show you whether Narratic fits.