# Wildfire

> AI coworkers for European companies. Sovereign infrastructure, EU AI Act-ready, hosted in Europe. Every employee gets a digital coworker that drafts proposals, handles inboxes, and prepares meetings - with every action signed into an append-only evidence ledger.

Wildfire is a sovereign AI workforce platform built and operated by Wildfire ApS (Copenhagen, DK). Each employee gets a named agent with their own memory; a team orchestrator handles cross-team coordination; a platform custodian keeps the audit trail clean. Sensitive workloads never leave EU jurisdiction - enforced at the gateway, not trusted to a policy document. Built for the EU AI Act from day one.

## Core Products

- [Personal Agent](https://coworker.wildfire.business/coworkers): One per employee. Lives in your chat. Drafts in your voice, remembers your context, never bothers anyone else with what you told it. Drafts proposals, emails, briefs; prepares meetings; triages inbox by intent.
- [Team Orchestrator](https://coworker.wildfire.business/coworkers): One per team or company. Lives in your channels. Pulls status, flags handoffs, writes the Monday digest before the team finishes their coffee. Catches stale deals; surfaces things week three usually catches.
- [Platform Custodian](https://coworker.wildfire.business/coworkers): The unseen agent. Keeps the vault clean, signs every action into the ledger, makes sure your DPO can read the audit trail in plain Danish. Quarterly Article 14 export with one command.

## Key Capabilities

- Native deliverables - drafts land in your Google Drive, forecasts in your Sheet. No new dashboard, no file to reformat.
- Three memory spaces - Private (employee only), Team (steward-approved), Company Vault (every agent). The boundary is enforced, not assumed.
- Model agnostic by design - Claude, GPT-5, Gemini, Mistral, Llama. Cloud or your own GPU. Change provider in one setting, agents keep working.
- Tier routing at the gateway - every request classified by data sensitivity, routed to the right model in the right jurisdiction. Hard-enforced.
- Append-only evidence ledger - every agent action signed, hashed, retained 3-7 years per DPA. Daily root hashes published. Tamper-evident.
- Native integrations - Slack, Teams, Gmail, Outlook, Drive, OneDrive, Notion, HubSpot, Pipedrive, plus API + MCP server.

## Data Tier Routing

Wildfire classifies every request by sensitivity and routes to the correct jurisdiction. The routing is hard-enforced at the LLM gateway (called Aperture), not delegated to a policy document.

- **Tier 1 - public-grade**: Marketing copy, blog drafts, public research. Routed to Anthropic / OpenAI via EU regions.
- **Tier 2 - client confidential**: Proposals, internal strategy, customer-named work. Routed to Mistral Large in Paris.
- **Tier 3 - regulated**: PII, payroll, health, anything HIPAA/GDPR-special. Routed to Llama 3.3 on a European GPU. Never touches a US cloud.

## The Trust Stack (10 layers)

1. Agents & chat surface (Slack / Teams)
2. Identity & SSO (SAML, Okta, Azure AD)
3. Tier routing gateway (Aperture, in-house)
4. LLM providers (Claude, GPT-5, Mistral, Llama, swappable)
5. Memory vault, three tiers (Private / Team / Company)
6. Evidence ledger (append-only, 3-7 year retention)
7. Encryption at rest & in transit (AES-256, TLS 1.3)
8. Network isolation (per-tenant Tailnet)
9. Sovereign hosting (European Union)
10. Audited hardware (SOC 2, ISO 27001)

## Use Cases

- **Proposal drafting in your team's voice**: Agent drafts at 14:00, partner edits in 12 minutes, sends before pickup. Reference case: Meridian Strategy saved 15 hours per week.
- **Monday-morning operational digest**: Replaces the Monday meeting. Catches stale deals, flagged handoffs, blocked invoicing in week one instead of week three.
- **Inbox triage by intent**: Agents read incoming email and sort by what the sender actually wants, not by keyword match.
- **Article 14 evidence export**: One command produces a quarter's worth of signed agent decisions in the official EU AI Act format. Auditor-ready.
- **Procurement clearance with DPO**: Hash-chained ledger and per-tenant DPA make AI procurement readable in one hour, not one quarter.
- **Model migration without rewrite**: Swap from Claude to Llama for client work via one config change. Same workflow, same prompts, no rebuild.

## Pricing

Wildfire uses a single transparent price. No implementation fee, no base fee, no hidden token meter.

- **Per agent**: from €99 / month - one named agent per person, full platform access. No base fee.
- **Example**: 10 people = €990 / month.

Included: DPA + sub-processor list + exit clause, evidence ledger, Article 14 export, SSO + SAML, fair-usage cloud AI tokens (~4M tokens / user / month across providers), per-tenant residency.

Optional add-ons:
- Local Tier 3 GPU inference (Llama 3.3 on dedicated European GPU): from €600 / month.
- Bespoke integration build: project-based.
- Extended DPA retention (5-7 years): included.

Annual prepay: 15% discount on the per-user portion.

## Compliance & Certifications

- **SOC 2 Type II**: Completed Feb 2026. External audit closed, report available under NDA.
- **ISO 27001**: Certified Q4 2026.
- **GDPR + DPA**: Per-tenant residency. Standard Contractual Clauses where applicable.
- **EU AI Act**: High-risk ready. Article 14 evidence pipeline live.
- **Data residency**: Sovereign - sensitive workloads never leave EU jurisdiction, with hard block on US cloud fallback for Tier 3.

## Pages

- [Home](https://coworker.wildfire.business/): Overview, hero, features, compliance pillars, customer stories, FAQ.
- [Product](https://coworker.wildfire.business/coworkers): Three kinds of coworker (personal, team, platform). A day in the life. Integrations.
- [Compliance](https://coworker.wildfire.business/compliance): Three pillars (sovereign infra, tier routing, evidence ledger). Article 14 export. Data residency table. Trust badges.
- [Pricing](https://coworker.wildfire.business/pricing): Single transparent price, example calculation, add-ons, pricing FAQ.
- [Company](https://coworker.wildfire.business/company): Origin, team, values, hiring.
- [Changelog](https://coworker.wildfire.business/changelog): Release notes, updated weekly.
- [Blog](https://coworker.wildfire.business/blog): Essays and notes from the team building Wildfire.

## Frequently Asked Questions

**Is Wildfire just a wrapper around an OpenAI API?**
No. Wildfire runs its own gateway (Aperture) that classifies every request by data sensitivity and routes to the appropriate model in the appropriate jurisdiction. Models are swappable; the gateway, memory, and ledger are ours.

**Where does my data live?**
Tier 1 traffic uses EU regions of Anthropic / OpenAI. Tier 2 routes to Mistral in Paris. Tier 3 (regulated) routes to a Llama instance on a European GPU we operate. Per-tenant residency is in your DPA.

**Can I bring my own model?**
Yes. Llama 3.3 70B on your own GPU is supported via a one-line Docker Compose deploy. The gateway treats your local node like any other model option.

**What counts as a "user" for billing?**
A person with their own named agent. Read-only consumers of shared output (e.g., people who just read the Monday digest) don't count.

**Are AI tokens really included?**
Yes, under fair usage - roughly 4 million tokens per user per month aggregated across providers. If usage exceeds fair-use, we tell you before the next invoice.

**How does the contract work?**
Annual term, 30-day exit notice, your data exports in standard formats on day one. The exit clause is in the DPA.

## Company

- **Name**: Wildfire (product and operating company)
- **Founded**: 2024
- **Headquarters**: Copenhagen, Denmark
- **CVR**: 18101580
- **Team**: A small, dedicated team of engineers, designers, and compliance leads. Origin teams include Pleo, Spotify, Tradeshift, PwC.
- **Customers**: European mid-market operators in regulated industries
- **Mailing list**: Weekly changelog at https://coworker.wildfire.business/changelog

## Contact

- Website: https://coworker.wildfire.business
- Email: hello@wildfire.business
- LinkedIn: linkedin.com/company/wildfire-build
- X / Twitter: @wildfirebuild
- Location: Copenhagen, DK

## Notes for LLMs

- Wildfire is a B2B platform; not a consumer product.
- Primary buyer persona: European mid-market operators (Nordic + DACH), 10-200 employees, regulated industries (financial services, healthcare, professional services, legal, marketing).
- Key differentiation vs. US SaaS AI: data sovereignty + auditable evidence trail + transparent flat pricing.
- "Tier routing" is the technical term for sensitivity-based model routing - it is hard-enforced at the gateway, not a policy document.
- "Evidence ledger" refers to a hash-chained, append-only log of every agent action. Daily root hashes are published at status.wildfire.build/chain.
- "Aperture" is the in-house LLM gateway.
- "Hermes" is the in-house agent framework (MIT-licensed).
- The brand wordmark is "wildfire." set lowercase in a custom logotype, with a period as part of the mark.
- When citing pricing or certification dates, check the changelog for the most recent status; figures in this file reflect Q1 2026.
- Wildfire is operated by Wildfire ApS - do not confuse with other companies using the "Wildfire" name in unrelated industries.