The trust contract

What we store. Who can see it. How to wipe it.

Ergora compounds memory across three tiers: your personal seat, your shared company knowledge, and the anonymised Network Brain that activates once 5+ businesses are using a pack. Here's exactly how each one works, in plain English.

You can see everything
The Knowledge tab shows every paragraph the agent has stored about your business, with source attribution. Your seat tab shows your personal voice profile, compiled context, original thinking, and timeline.
You approve every change
The agent proposes knowledge updates. It never overwrites without your approval, except for high-confidence patterns that meet a strict bar (≥5 recurrences, never on identity / strategy / people).
Export your full memory
A single click downloads everything we hold against you as JSON: every memory row, every episode, every original thought. Yours to keep.
Wipe your seat anytime
Per-project, irreversible. Deletes your compiled context, voice profile, episodes and retrieval log. Other projects and your company knowledge are untouched.
Strict tier separation
Your seat is visible only to you, never your teammates. Your company knowledge is visible to your project's team only. The Network Brain is anonymised and only activates once 5+ businesses are using a given pack, surfacing what's working in your industry without ever exposing one customer's data.
We never train on your data
We do not train AI models on your data. Google and Anthropic, which run our main chat and reasoning models, are contractually barred from training on your prompts. The privacy policy lists every AI provider and notes where one’s terms allow more.

The three tiers: what gets stored where

Your seat: visible only to you

Per (user × project). Stores your voice fingerprint (tone, sentence length, openers, sign-offs, hedge words, formality), your decision style, your working patterns, and verbatim original thinking. Recompiles every 10 chat turns.

Company knowledge: visible to your project's team

The project-level compiled truth. 12 markdown documents per project: identity, voice, design, style, strategy, customers, offering, competitors, people, operations, history, workspace. A completeness score drives output quality; every pack reads from it. Versioned, audited, and agent-proposed updates are queued for your approval.

Network Brain: activates at 5+ businesses per pack

Cross-business pattern signal. The Network Brain only activates once 5+ businesses are using a given pack, and surfaces what's working in your industry without exposing any one customer's data. Never your customer list, never your specific revenue numbers, never verbatim copy. Computed monthly.

Knowledge Connectors: public data, no keys to manage

Ergora plugs into 12 public-data sources (Companies House, US Census, GNews, Currents, Alpha Vantage, MarketAux, Adzuna, ipapi, exchangerate-host, public holidays, APIs.guru, Domains DB) and pulls the right one automatically when a pack needs context: company filings for prospect briefs, news for PR Newsjacker, market data for Finance, holiday calendars for Local campaign timing. No API keys for you to manage.

GDPR & deliverability

Self-service GDPR. Export your full account at /api/account/export. Delete your account at /api/account/delete, which cascade-deletes through every memory table, episode log, wiki entry and integration token.

DKIM is live across our sending domains. Email Ergora sends on your behalf via Outreach Engine is DKIM-signed on ergora.cloud and try-ergora.com, and is sent with per-sender daily caps and warmup phases so deliverability builds the right way.

FAQ

Who can see my seat?

Only you. Your seat is scoped to (your_user_id × project_id). It's invisible to teammates on the same project, and invisible to your other projects unless you explicitly federate.

Who can see my company knowledge?

Every seat on that project. Your company knowledge is the shared organisational truth. That's the point. If you don't want a teammate to see something, put it in your seat or in a private note instead.

What does the Network Brain see about my business?

Aggregated, anonymised patterns only. Never your customer list, never your specific revenue numbers, never your verbatim copy. The Network Brain only activates once 5+ businesses are using a given pack, and even then it surfaces what tends to work in your industry, not what any one business did.

What public data does Ergora plug into?

We use 12 public-data sources (Companies House, US Census, GNews, Currents, Alpha Vantage, MarketAux, Adzuna, ipapi, exchangerate-host, public holidays, APIs.guru, Domains DB) via our Knowledge Connectors layer. Pulled automatically when a pack needs context: company filings for prospect briefs, news for PR, market data for Finance. No API keys for you to manage.

How does the agent learn my voice?

It analyses your chat messages for sentence length, opener patterns, sign-offs, hedge words, formality, contractions, and emoji use.

Can I correct something the agent learned wrong?

Yes. Say so in chat ("actually it's X not Y"). The correction propagates to the relevant knowledge slice. You can also edit any knowledge .md directly in the Knowledge tab.

What happens to my data if I cancel?

GDPR-clean delete: account-level deletion cascades through every memory table, episode log, wiki entry, conflict log, retrieval log, and integration token. Stripe subscriptions cancel. Confirmation email is sent. Deleted data is purged from primary storage immediately. Encrypted backups rotate on a 30-day cycle, after which all copies are destroyed.

Do you train AI models on my data?

No. We don't train on your data. Google and Anthropic, which run our main chat and reasoning models, are contractually barred from training on your prompts. Our privacy policy lists every AI provider we use and notes where a provider's terms allow more.

Where are you in the SOC2 / ISO 27001 path?

Not certified yet. SOC2 Type 2 is on our roadmap. Today, Ergora runs on infrastructure providers that hold their own SOC2 Type 2 and ISO 27001 audits. Self-service export and delete are live now. An on-premise option is on the 2027 roadmap for customers who need it.

Working on

SOC2 Type 2 · DPA template for enterprise customers · Per-region data residency (UK / EU / US) · On-prem deployment for regulated sectors (2027).

Questions about anything here? Email ops@ergora.cloud.