30–60 minutes per ticket

Gathering context scattered across Jira, Slack, Confluence, GitHub, and meeting transcripts.

Your context, compiled.

Decisions, contradictions, and open questions — surfaced from every source.

Local-first intelligence for engineering teams

Enter a ticket. Get a dossier.

The full picture on any ticket, before your agent touches the code.

Request Early Access See how it works

Currently in development

Scroll
Why Dossier
6 Slack threads.
3 Confluence pages.
2 open PRs.
47 Jira comments.
1 stale spec. (no one updated it)
1 brief.
Local-first — data never leaves your machine
Cross-source synthesis — not summaries, reconciliation
Contradictions flagged — stale docs surfaced automatically
Structured briefs — decisions, people, risks, all attributed
BYOK AI — your key, your model, no lock-in
No admin required — download, connect, go
r.okonjo 2:47 PM
honestly I thought we settled this in the retro — Postgres was the call
In Review MIGR-12
AssigneeF. Holmström
SprintSprint 14
Points5
PR #482 Replace Redis session store
const store = createStore({
- adapter: redisAdapter(client),
+ adapter: pgAdapter(pool),
});
In Review
MIGR-12
Redis to Postgres Migration
Summary
Team aligned on Postgres migration for session storage. Confluence spec still references the original Redis approach and needs updating.
Timeline 4 items
Key People 3
Contradictions 1
Contradiction detected
Confluence spec describes Redis — team decided Postgres
How it works
01

Enter a ticket

Paste a ticket ID or URL. Dossier identifies the source and starts gathering.

Jira MIGR-12 In Review
Migrate session store from Redis to Postgres
Sprint 14 3 subtasks
02

Extracted & reconciled

The LLM reads every source, extracts structured facts, and reconciles what conflicts.

Raw signals
"let's not add Redis for this"
Redis-backed queue with Bull
Status: In Progress
PR #412: skip locked impl
→
Structured output
Decision: Postgres SKIP LOCKED
Contradiction: spec says Redis
Gap: spec not updated
03

Brief compiled

A structured intelligence brief with a decision trail, key people, code changes, and gaps identified.

DOSSIER
BRIEF-2026-0314
COMPILED
RE: MIGR-12 — Migrate session store from Redis to Postgres
Summary
Decision Trail 8 events
Key People 4
Code Changes 2 PRs
Open Questions 2
Contradictions 1
5 sources 32 signals compiled 4s ago
What you get

A structured intelligence brief.

Not a summary. Not a chat log. A structured brief with a decision trail, key people, contradictions flagged, and open questions surfaced — all attributed to sources.

Dossier — NOTIF-7
Summary
Approach
Timeline
People
Questions
Contradictions
Blockers
Reading
Generated just now · 24,588 tokens · 87.0s
Jira GitHub
NOTIF-7
WebSocket Notification Migration
Jira ✓ GitHub ✓
IN REVIEW
PriorityHigh ReporterBex Hargrove AssigneeTomás Nkedi SprintSprint 4
Summary

PR #247 approved, blocked on load testing. Team aligned on 30-day parallel operation with infrastructure budget approved.

Current Approach

Feature-flagged rollout with automatic fallback to polling if WebSocket connections drop below health threshold.

Decision Trail
Mar 6
Ticket created for migrating notifications from polling to WebSocket
Jira - NOTIF-7 →
Mar 9
Infrastructure budget approved — $1,200/month for additional server nodes
Jira - NOTIF-7 →
Mar 12
Agreed to 30-day parallel operation with feature flag capability
Jira - NOTIF-7 →
Frequently asked

Where does my data go?

Nowhere. Dossier runs locally on your machine. Source data is fetched directly from your accounts via OAuth, processed locally, and stored in a SQLite database on disk. Nothing is sent to any server we operate.

What LLM does it use?

Whichever you choose. Bring your own API key for Claude, GPT-4, or any OpenAI-compatible provider. Your key, your model, your cost structure. We never see your API key.

What does it cost?

Dossier is free. You pay only for your own LLM API usage, which typically runs $0.02-0.08 per brief depending on source volume and model choice.

What sources are supported?

Jira, Slack, Confluence, GitHub, and Granola (meeting transcripts). Linear support is planned. Each source connects via OAuth with read-only permissions.

Do I need admin permissions?

No. Dossier uses your personal OAuth tokens with the same access you already have. No IT approval, no org-wide deployment, no SSO configuration.

How long does a brief take?

Typically 30-90 seconds depending on source volume and LLM speed. The first brief for a ticket takes longest as sources are fetched; subsequent regenerations are faster.

No vendor sales call. No org-wide deployment. No admin permissions needed. Download it, connect your accounts, run your first brief in under a minute.

Pricing

Start free. Scale when you're ready.

Save $24/year
Free · 5 briefs/mo · All connectors · BYOK · Get started
Pro Recommended

Everything, unlimited.

$ 12 /month
Billed annually at $120/year
  • Unlimited briefs
  • All source connectors
  • BYOK — bring your own key
  • Priority support
Get Started

Enter a ticket.
Get a dossier.

We're building Dossier in the open. Request early access to be first in line.

Request Early Access

macOS app. Local-first. Currently in development.