# Entradigm — agent-native marketing-automation platform

Entradigm is built FOR AI agents first, with a thin human UI on top. Every UI
action has a schema-typed API + (roadmap) MCP equivalent. This file is the
machine-readable entry point for agents discovering the platform.

## Domain model

- tenant   — an isolated brand workspace. All data is row-level isolated; one
             agent/operator acts within exactly one tenant at a time.
- profile  — the tenant config: brand voice, content rules, schedule, the BYO
             provider routing (AI/TTS/music/social). Credentials are YOURS —
             the platform never holds raw provider keys.
- cycle    — one end-to-end content run (ideate → render → review → publish).
             Triggered via POST /strategist/run or the manual trigger.
- signal   — an external input (trend / event / source) that can seed a cycle.
- HITL     — human-in-the-loop review. A cycle pauses at "pending_review"; a
             human (or an authorised agent) approves or parks it before publish.

## Platform documentation

<<<ENTRADIGM-DOCS-START>>>

## Who it's for (start here)
Three kinds of operator hire Entradigm for three different jobs. Find yours, then follow the trail.

### Individual / creator — "Keep me visible without it eating my week"
_Solo founder, freelancer or creator who needs a consistent presence without an agency._

**The job you're hiring us for**: show up consistently on social without spending your week on it.

If you're a solo founder, freelancer or creator, the wall isn't ideas — it's the production cost of turning them into posts, day after day. Entradigm runs that loop for you: an agent drafts and renders each cycle, you approve it from your phone in a minute.

What to do next:

- Read [Entradigm in one sentence](#a-what-is-overview) for the mental model
- See the real numbers in [Cost of Entradigm](#a-cost-roi-cost-entradigm) — typically €15-€60/month all-in at daily cadence
- In onboarding, pick the **Individual / creator** persona on step 1 — the wizard relaxes the audience question and uses creator-flavoured examples

### Small business — "Punch above our weight without a marketing hire"
_A small team with no dedicated marketing department that still needs a credible, consistent brand._

**The job you're hiring us for**: a credible, consistent brand presence without hiring a marketing function.

Small teams know cadence matters but can't spare 8-12 hours a week on content. Entradigm collapses that to a few minutes of review per cycle, and keeps the whole pipeline coherent — brand voice, content rules and history all live in one place, so three months in your feed still reads like one brand, not five.

What to do next:

- Read [Why fragmented tools produce fragmented brands](#a-why-you-need-narrative-coherence)
- Weigh it against [the cost of doing nothing](#a-cost-roi-cost-nothing)
- In onboarding, pick the **Small business** persona on step 1

### Mid-market — "Scale output and keep compliance airtight"
_An established company with a marketing function that needs more output without losing review rigor or compliance._

**The job you're hiring us for**: scale content output across channels while keeping review rigor and compliance airtight.

If you already have a marketing function, the constraint is throughput plus governance — especially in regulated sectors (telecom, finance, health). Entradigm adds capacity without loosening control: multi-stage review, per-tenant content rules, full provenance, and EU AI Act Art. 50 labelling on every AI-assisted surface.

What to do next:

- Read [Multi-stage review for regulated industries](#a-features-multi-stage-review)
- Review the [compliance & security](#c-compliance) category in full
- In onboarding, pick the **Mid-market** persona on step 1 (the default)


## What is Entradigm?
Start here. The shortest possible explanation of what we built and why it exists.

### Entradigm in one sentence
_The agent-native automation platform — software built for AI agents, with a thin human UI on top._

Entradigm is the threshold to the new paradigm where AI agents are the primary citizens of software, and humans interface through a thin layer.

The name comes from Latin *intrare* ("to enter / cross the threshold") + Greek δείκνυμι *deiknumi* ("to show / to model") — the root of *paradigm*.

What this means in practice:

- Every feature exposes a schema-typed API so an agent can drive it directly
- Every workflow is a declarative intent that an agent can compose
- Every credential stays in your hands — we never hold your OAuth tokens, API keys, or social-account passwords
- The human UI is small on purpose — it's the steering wheel, not the engine

### The mental model — one worker, many tenants, your stack
_How Entradigm runs your content while never touching your provider credentials._

One Cloudflare Worker orchestrates every cycle. Per-tenant configuration lives in our database, per-tenant CREDENTIALS live in your accounts.

When a cycle fires for your tenant, the Worker:

1. Reads your tenant config (brand rules, content rules, schedule, AI provider, social accounts)
2. Calls YOUR AI provider with YOUR key (we never see your bill)
3. Renders the carousel + audio + captions using YOUR TTS + music providers
4. Uploads to YOUR Blotato account → posts to YOUR social handles
5. Sends the review link to YOUR Telegram bot → you approve

We hold: tenant config, content drafts, audit logs.
We never hold: Anthropic / OpenAI / ElevenLabs / Suno / Blotato / Telegram credentials.

### Why now: the agent-native shift
_The next billion users of the internet won't be humans — they'll be AI agents._

Software built for humans first is reaching the end of its lifecycle. The next billion users of the internet will be AI agents acting on behalf of humans — booking, buying, posting, monitoring, deciding.

We built Entradigm for that future, not for the present.

The platform's primary surface is its declarative API + MCP-callable runtime. The human UI is the second surface, layered on top. This is the inverse of most SaaS: most tools have a human UI with an API bolted on. Entradigm has an agent surface with a human UI bolted on.

Practical consequence: every action you can do in the UI, an agent can do via the API. We don't have "internal-only" endpoints.


## Why you need it
The case for adopting Entradigm now, not later.

### The cadence gap that's eating your reach
_Daily posting beats weekly. Weekly beats monthly. Most teams post when someone remembers — that's the gap._

Social algorithms reward cadence. The accounts that grow are the ones that post consistently — daily, or close to it. The accounts that stagnate are the ones that post when someone has time to remember.

Teams know this. The reason they don't post daily isn't lack of will — it's that the *production cost* of each post is too high:

- 30-60 min ideation
- 30-60 min copy + design
- 15-30 min review + scheduling
- 5-10 min posting + cross-posting

That's 1.5-2.5 hours per post × daily × 5 channels = 7.5-12.5 hours / week of pure production time. Most teams have 0-2 hours / week to spare.

Entradigm collapses this to 5 minutes / day of human review. The strategist agent does the rest.

### Why fragmented tools produce fragmented brands
_A separate ideation tool + a separate copy tool + a separate scheduler = three brands._

Most teams stitch together: a brainstorming tool (or Notion), a copy generator (Jasper / Claude direct), a design tool (Canva / Figma), a scheduler (Buffer / Hootsuite), and a review surface (Slack threads).

Each tool optimizes for its own slice. None of them sees the cumulative brand. Three months in, your feed reads like five different brands in a trenchcoat.

Entradigm holds the WHOLE pipeline. Brand voice + visual identity + content rules + cycle history all live in one place. The strategist sees what you posted yesterday before drafting today. Coherence isn't a goal you write down — it's a property of the system.

### Why EU-jurisdiction matters in 2026
_Data residency, EU AI Act, GDPR — choose where your data lives, with whom._

2025-2026 ratcheted regulatory exposure for any business publishing content in the EU. EU AI Act Article 50 requires AI-assisted content to be labelled. GDPR has been around for years, but enforcement is getting sharper.

Entradigm runs on Cloudflare (EU-region available) + Supabase (EU-region available) + your choice of LLM provider. If you choose Anthropic EU or OpenAI EU, your content data never leaves EU jurisdiction.

We surface the sub-processor map in your settings → you can see, at any moment, every third-party that touches your data. You approve each one explicitly.


## Cost & ROI
The three numbers that matter: cost of doing it manually, cost of doing nothing, cost of doing it with Entradigm.

### Cost of doing it manually
_~10 hours / week × your loaded hourly rate. Calculate yours._

Conservative estimate for a marketing team running daily content across 5 platforms:

- Ideation: 1 hour / day
- Copy + design: 1 hour / day
- Review + revise: 30 min / day
- Schedule + post: 30 min / day
- **Total: ~3 hours / day = ~15 hours / week**

At a loaded hourly rate of €60-€120 (which is the actual cost including overhead, not the wage), this is **€900-€1,800 / week = €46k-€94k / year**.

That's the floor. It assumes the work actually gets done at that cadence — which most teams can't sustain.

### Cost of doing nothing
_Algorithmic reach decay + sales atrophy + brand drift compound silently._

The expensive scenario isn't "spend €50k/yr on a marketing team." It's "spend €0 on marketing, lose €200k/yr in pipeline you would have had."

Three compounding costs:

- **Algorithmic decay**: every social platform amortizes follower-reach to accounts that post. Skip a week → reach drops 20-40%. Skip a month → reach drops 60-80% and recovers slowly. The opportunity cost of missing the cadence window compounds.
- **Sales pipeline atrophy**: B2B SaaS conversion data shows 6-12 inbound touchpoints typical before a buyer reaches out. Each missed touchpoint = a buyer who never enters the funnel.
- **Brand drift**: without consistent voice + visual identity reinforcement, you become "that company we vaguely remember" instead of "the company that solves X." Vague memory doesn't close deals.

These costs don't appear on any expense line. They appear as the pipeline that never materialized.

### Cost of Entradigm — per-cycle anchor + BYO LLM
_~€0.15/cycle infrastructure + your AI-provider bill (typically €5-€50/month at daily cadence)._

Entradigm's cost has two layers:

**Infrastructure cost** (paid to ByKA Insights): ~€0.15 / cycle, billed monthly. Daily cadence = ~€4.50 / month / tenant.

**AI provider cost** (paid by you, to your provider): depends on the LLM you choose. At daily cadence:

- Anthropic Claude Sonnet: ~€10-€20 / month
- Anthropic Claude Haiku: ~€2-€5 / month
- OpenAI GPT-4o: ~€8-€15 / month
- Your locally-hosted open model: €0 + your compute cost

You see your AI bill on your Anthropic / OpenAI dashboard — we never see it, never mark it up.

**Total typical**: €15-€60 / month / tenant for daily cadence across 5+ platforms.

Compare to the €4k-€8k / month manual cost from "Cost of doing it manually."

### Typical payback period
_Most tenants recover the full Entradigm cost within 2-4 weeks of the first published cycle._

Payback math for a typical B2B SaaS tenant:

- Manual cost saved: ~15 hrs/wk × €80/hr = €1,200/wk = €4,800/month
- Entradigm cost: ~€30/month (€0.15 × 30 cycles + AI provider)
- Net saving: ~€4,770/month
- Payback: ~5 hours into the first week

Even if your team's hourly cost is much lower (€20-€30/hr), payback is under 2 weeks.

This calculation doesn't include the upside — the pipeline you build by being consistent. That's harder to quantify but usually larger than the cost saving.


## What is unique about Entradigm
Four properties no incumbent has. These are the load-bearing differentiators.

### BYO-everything posture
_You bring your own AI / TTS / music / Telegram / Blotato / notification credentials. We never hold them._

Most marketing platforms require you to give them your social account passwords, your OpenAI key, your TTS subscription. They become a credential broker — every credential is one breach away from compromise.

We never hold raw credentials. The setup flow:

- You buy your own Anthropic / OpenAI subscription → you control the key
- You buy your own ElevenLabs / Suno / Mubert subscription → you control the key
- You buy your own Blotato subscription → Blotato holds your social tokens, not us
- You create your own Telegram bot → its token belongs to you
- You define your own webhook destinations for notifications

Per-tenant config tells the Worker which BYO providers to call for each cycle. The provider call goes out FROM the Worker, but the credentials are passed in from your encrypted-at-rest tenant config — never logged, never echoed, never stored in plaintext.

**Why this matters**: cost ownership stays with you, license-chain audit stays clean, breach exposure is your responsibility not ours, and you can move off Entradigm tomorrow without losing your accounts.

### Agent-callable by design
_Every UI action has an API + MCP equivalent. We don't have "internal-only" endpoints._

The Build FOR AI thesis (vs. WITH AI) means the agent surface is the primary surface, not the bolted-on one. Concretely:

- Every cycle trigger has both a UI button and a POST /strategist/run API endpoint
- Every brand-rule edit has a UI form and a PATCH /api/tenants/{id}/brand endpoint
- The whole platform is wrappable as an MCP server — agents can use it as a tool
- Schemas are introspectable: [GET /strategist/schema](/strategist/schema) returns the JSON schema of every intent

Two live, public, zero-auth discovery surfaces let an agent learn the platform before it touches a single credential:

- [GET /strategist/schema](/strategist/schema) — the machine-readable JSON schema of every intent + the platform capabilities block (apiVersion, capabilities)
- [GET /.well-known/llm.txt](/.well-known/llm.txt) — a plain-text orientation file for LLM agents (also served at [/llms.txt](/llms.txt) and [/llm.txt](/llm.txt)): the platform docs flattened to markdown + the live endpoint list + copy-paste curl recipes

Fetch the schema in one line:

```bash
curl -s https://entradigm.com/strategist/schema | jq '.capabilities'
```

An external agent (your n8n workflow, your Make scenario, your Devin / Cognition agent) can drive Entradigm exactly the same way a human can.

### EU-jurisdiction first
_Default deployment region is EU. US is opt-in. Sub-processor list is in your dashboard._

We're a Bavaria-based operator. Our customers' first market is usually Germany / DACH / EU. EU jurisdiction is the default, not the afterthought.

Cloudflare Workers run in the EU region for EU tenants. Supabase EU region holds the database. Your AI provider can be Anthropic EU or OpenAI EU.

The sub-processor list (Settings → Privacy → Sub-processors) shows every third-party that touches your data + which region they operate in + the DPA reference. You explicitly approve each one.

### Open architecture, escapable
_Your data stays exportable. Your providers stay swappable. No lock-in by design._

Lock-in is how SaaS companies extract rent. We chose the opposite posture:

- Your data is exportable via GDPR Art. 20 portability endpoint (JSON + CSV) — anytime, no friction
- Your AI provider is swappable — change the key in Settings, the next cycle uses the new provider
- Your social accounts are swappable — they're in your Blotato account, not ours
- The platform itself runs on commodity infrastructure (Cloudflare + Supabase) you could replicate

If you decide to leave Entradigm, you take everything that's yours and we wave goodbye. The system is designed to be escapable — which keeps us honest about delivering value every month.


## Features
What the platform actually does. Grouped by the lifecycle of a cycle.

### Multi-tenant brand isolation
_One platform, N tenants. Brand A and Brand B never see each other._

Each tenant has its own brand rules, content rules, social accounts, AI provider config, schedule, and review history. Tenant data is row-level isolated in Postgres (Supabase RLS) — even an agency operator with access to multiple tenants sees only the one they're switched into.

Pricing scales per-tenant, so the multi-tenant feature is free up to your plan's tenant count.

### Cycle orchestration — strategist → reviewer → publisher loop
_The three-agent loop that runs every cycle._

Every cycle is three coordinated agents:

- **Strategist** — generates the content draft (carousel slides + caption + alt-text)
- **Reviewer** — validates against brand rules + content rules + anti-pattern list + compliance gates
- **Publisher** — once approved, posts to each social platform via Blotato + sends review link via Telegram

The loop runs end-to-end in 2-5 minutes. You see the result as a Telegram review link with APPROVE / PARK buttons.

### Cycle scheduling — cron + auto-approve thresholds
_Cron expressions for cadence; per-tenant blackouts; auto-approve when confidence > threshold._

Per-tenant schedule supports:

- Standard cadences (daily / MWF / weekly / monthly)
- Custom cron expressions for non-standard windows
- Blackout windows (holidays, conferences, sensitive dates)
- Per-cycle confidence thresholds — above X, auto-approve; below, human review

Time zone is per-tenant (defaults to operator's TZ). Multi-tenant operators see all schedules in one calendar view.

### Brand derivation — auto-fill from your materials
_Point at your website + brand book + past posts → AI drafts the brand profile._

Instead of filling 30 form fields manually, paste your website URL + optionally upload a brand book PDF + a few past social posts.

The brand-derive agent reads them and proposes a brand profile: voice keywords, anti-patterns, visual identity inputs (colors, fonts), tone register, target audience phrasing.

You review the draft + adjust + save. Typical onboarding time goes from 45 minutes to 8 minutes.

### Multi-stage review for regulated industries
_Strategist → Legal-reviewer → Approver routes for telecom / finance / health._

Default review = one human approval. Regulated tenants can add stages:

- Strategist drafts → Legal-reviewer agent runs compliance check (claims, disclaimers, AI-disclosure cadence) → Human approver
- Optional human legal reviewer slot for high-stakes posts
- Per-stage Telegram routing — legal reviewer gets a different chat than the approver

Configurable per-tenant. Once set, every cycle routes through the configured stages.

### Render pipeline — Remotion + R2 + Blotato
_Carousel slides rendered as videos, hosted in R2, posted via Blotato._

Carousel content renders through Remotion (React-based video framework) into MP4 + still frames. Assets upload to your Cloudflare R2 bucket (BYO bucket optional; default = ours).

Blotato handles cross-platform posting (LinkedIn / X / Instagram / Facebook / TikTok / YouTube Shorts). The Worker calls Blotato with the post payload + asset URL; Blotato handles platform-specific quirks.


## How to onboard
From "I bought this" to "first cycle published" in 15 minutes.

### Sign up + first sign-in
_Invitation-only. Passwordless magic-link sign-in + mandatory authenticator-app (TOTP) two-factor._

Entradigm is invitation-only — there is no public self-signup. An operator invites you by email; the invite is a single-use link valid 7 days.

When you accept, you sign in passwordlessly: we email you a one-time magic link, you click it, then you set up a mandatory authenticator-app (TOTP) second factor and save your recovery codes. After that you land in the onboarding wizard. We never hold a password — there isn't one to leak.

If you're an agency operator and want to add additional tenants later, do it through Settings → Tenants → New tenant.

### The 7-step wizard walkthrough
_Brand → voice → visual → sources → schedule → notifications → finalize. 8-12 minutes typical._

The wizard walks through:

1. **Organization basics** — name, industry, primary language
2. **Brand voice** — keywords + anti-patterns
3. **Visual identity** — colors, fonts, carousel style preference
4. **Sources** — your website + brand materials (for brand-derive)
5. **Schedule + cost preview** — pick cadence, see the per-month cost preview
6. **Notifications** — Telegram bot setup (paste your bot token + chat ID)
7. **Review + finalize** — confirm everything, click Create

8-12 minutes typical, faster if you let brand-derive do step 2-3 for you.

### Your first cycle, end-to-end
_Trigger a cycle, watch the strategist work, approve the result._

After the wizard:

1. Click "Trigger cycle" on the cycles page → strategist starts
2. Watch the cycle progress (2-5 min)
3. Receive Telegram review link → open it
4. Approve or park with a note
5. If approved: Blotato posts to your social accounts; you see a success row in cycles list

After this first cycle, scheduling takes over — the cron fires per your config.

### Inviting team members
_Settings → Team → Invite. Three role tiers._

Three role tiers per tenant:

- **Operator** — can trigger cycles, approve, edit brand, edit schedule
- **Reviewer** — can approve / park cycles, cannot edit brand or schedule
- **Viewer** — read-only access to cycles + brand config

Invites go via email as a single-use link (valid 7 days). The invitee signs in passwordlessly via a magic link, sets up a mandatory authenticator-app (TOTP) second factor with recovery codes, then gets the role you assigned. You can revoke any role at any time.


## BYO — Bring Your Own (configure each)
Six BYO surfaces. Each is optional with a sensible default; each is configurable per-tenant.

### BYO AI provider — Anthropic / OpenAI / custom endpoint
_Default: ByKA-shared Anthropic. Switch to your own key anytime._

Settings → AI provider lets you choose:

- **ByKA shared** (default) — pay-as-you-go via your subscription
- **BYO Anthropic** — paste your Anthropic key (sk-ant-…); model selectable
- **BYO OpenAI** — paste your OpenAI key (sk-…); GPT-4o / GPT-4.1 / GPT-4 selectable
- **BYO custom endpoint** — any OpenAI-compatible endpoint URL + key (Ollama, vLLM, LiteLLM proxy, your locally-hosted Llama, etc.)

Keys are encrypted at rest in Supabase Vault, never displayed back, never logged. You can rotate any time.

### BYO TTS provider — ElevenLabs / OpenAI / Azure / custom
_For carousel narration. Optional; cycles work without narration too._

Settings → TTS provider lets you choose:

- **None** — skip narration (carousel posts as silent slides)
- **BYO ElevenLabs** — paste your ElevenLabs key + voice ID
- **BYO OpenAI TTS** — uses your OpenAI key (already set in AI provider)
- **BYO Azure Speech** — paste your Azure key + region
- **BYO custom endpoint** — any OpenAI TTS-compatible endpoint

License-chain stays clean: you hold the TTS subscription, you hold the voice license, we never store generated audio beyond the cycle's lifetime.

### BYO Music provider — Suno / Mubert / Soundraw / custom library
_Background music for carousel videos. License-chain stays in your control._

Settings → Music provider lets you choose:

- **None** — silent carousels (default)
- **BYO Suno** — paste your Suno API key (commercial-license tier)
- **BYO Mubert** — paste your Mubert API key
- **BYO Soundraw** — paste your Soundraw API key
- **BYO custom library** — upload royalty-free tracks to your own R2 bucket; we reference them by URL

We never re-distribute music. The license-chain is: provider → your account → your post. No middleman.

### BYO Telegram bot — for review links + cycle notifications
_You create the bot via @BotFather, paste its token, we never see chat history._

Telegram review links are how you approve / park cycles from your phone. To set up:

1. Open @BotFather in Telegram
2. /newbot → name + username → BotFather gives you a token
3. Paste the token in Settings → Notifications → Telegram
4. Send /start to your new bot → forward the message to us so we get the chat ID
5. Save

We send review links to that chat. We never read incoming messages other than your APPROVE / PARK button replies.

### BYO Blotato — social posting layer
_Blotato holds your social OAuth tokens. We don't._

Blotato is the social-posting middleware that holds your LinkedIn / X / Instagram / Facebook / TikTok / YouTube tokens. To set up:

1. Sign up for Blotato (their pricing page)
2. Connect your social accounts inside Blotato
3. Paste your Blotato API key in Settings → Social → Blotato
4. Per-network: paste the account ID from Blotato into the matching field

We call Blotato; Blotato calls the platforms. Your tokens never touch our infrastructure.

### BYO notification channels — Slack / email / webhook
_Beyond Telegram: route cycle events to where your team works._

Settings → Notifications → alternative channels:

- **Slack webhook** — paste your Slack incoming-webhook URL
- **Email** — set a destination email; we send via your SMTP or our Postmark fallback (BYO Postmark optional)
- **Generic HMAC-signed webhook** — paste a URL; we POST cycle events signed with a shared secret

You can route different events (cycle ready / cycle failed / cycle approved) to different channels.


## Compliance & security
GDPR, EU AI Act, sub-processors, encryption. The serious side.

### GDPR rights — Art. 15 / 17 / 20
_Data export, data erasure, data portability. 30-day SLA._

Self-service GDPR endpoints under Settings → Privacy:

- **Art. 15 (Access)** — download a JSON + PDF of every data point we hold about your tenant
- **Art. 17 (Erasure)** — request deletion; 30-day SLA; we cascade across all sub-processors
- **Art. 20 (Portability)** — export in JSON + CSV for migration to a competitor

For human review or legal questions: see Settings → Privacy → Contact DPO.

### EU AI Act Article 50 — AI-disclosure cadence
_How we label AI-assisted content + why the "C3 AI disclosure" consent cannot be disabled._

EU AI Act Article 50 (effective 2026) requires AI-generated or AI-substantially-assisted content to be labelled.

Our default: every cycle includes a footer disclosure like "AI-assisted, reviewed by [Operator name]" in the caption. The exact wording is configurable per-tenant, but the PRESENCE of the disclosure cannot be disabled. That's a regulatory floor, not a product choice.

If you operate in a non-EU jurisdiction, the disclosure is optional — but we keep it on by default because most ByKA tenants serve EU customers.

### Sub-processor list + SCC posture
_Anthropic / Supabase / Cloudflare / Exa / Blotato — what each touches._

Current sub-processors (visible in Settings → Privacy → Sub-processors):

- **Cloudflare** — Workers runtime + R2 storage + DNS (EU region available)
- **Supabase** — database + auth (EU region available)
- **Anthropic** — LLM (EU endpoint available; or BYO your own provider)
- **Exa** — web research (US-based; opt-in)
- **Blotato** — social posting (US-based; mandatory if posting to social)
- **OpenAI** — only if you choose BYO OpenAI for AI / TTS

Each has a Standard Contractual Clauses (SCC) reference + most have EU regions.

### Encryption + secret handling
_TLS everywhere; Vault-encrypted secrets; presence-only logging._

All in-transit traffic is TLS 1.3.

All secrets (BYO keys, tokens, OAuth refresh tokens that we issue ourselves) are encrypted at rest using Supabase Vault.

Our pre-tool tripwire blocks any command that would echo a secret value to logs or stdout. Logs contain only "present / absent" for secret-shaped values.

If a secret is accidentally exposed (operator mistake, log leak), our incident-response runbook includes immediate rotation across all sub-processors.


## Integrations
Every external service Entradigm talks to + how the connection is established — and how an external agent connects to Entradigm.

### Agent API & discovery surfaces
_How an external agent finds, introspects, and drives Entradigm — schema + llm.txt, no auth to read._

Entradigm exposes two public, zero-auth surfaces an agent reads to orient itself, plus the bearer-gated run endpoint it calls to act:

- [GET /strategist/schema](/strategist/schema) — JSON schema of every intent + the capabilities block (apiVersion + the list of what this deployment can do). CORS-enabled, readable from any origin.
- [GET /.well-known/llm.txt](/.well-known/llm.txt) — plain-text orientation for LLM agents; also at [/llms.txt](/llms.txt) and [/llm.txt](/llm.txt). Contains the docs flattened to markdown + the live endpoint list + curl recipes. PLATFORM docs + PUBLIC schema only — never any tenant data.
- POST /strategist/run — the action endpoint. Bearer-token gated + rate-limited; an agent supplies the tenant's token to trigger a cycle.

A typical discovery handshake:

```bash
# 1. orient
curl -s https://entradigm.com/.well-known/llm.txt

# 2. introspect the intent schema
curl -s https://entradigm.com/strategist/schema | jq '.intents'

# 3. act (bearer-gated)
curl -s -X POST https://entradigm.com/strategist/run \
  -H "Authorization: Bearer $ENTRADIGM_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"intent":"generate-cycle","tenant":"acme"}'
```

Every UI action has an API equivalent — there are no internal-only endpoints.

### Anthropic Claude — default LLM
_Claude 4.7 Sonnet for cycle strategist; Haiku for cheaper tasks._

Default model bindings:

- Strategist (content draft) — Claude Sonnet 4.7
- Reviewer (compliance + brand-rules check) — Claude Haiku 4.5 (cheaper, faster)
- Brand-derive — Claude Sonnet 4.7

You can override per-tenant via BYO Anthropic key + model choice.

### Blotato — social posting
_BYO Blotato required for posting to LinkedIn / X / Instagram / etc._

Blotato is the only mandatory third-party (if you want to post to social). Why: holding social OAuth tokens has compliance complexity we chose not to take on. Blotato specializes in it.

Setup: see "BYO Blotato" in the BYO category.

### Telegram — review + notifications
_BYO Telegram bot for cycle review links + alerts._

Telegram is our default review surface because it's mobile-native + low-friction for approval buttons.

Alternative channels (Slack / email / webhook) are configurable; see BYO Notifications.

### Cloudflare — Worker runtime + R2 storage
_Workers run the orchestration; R2 holds rendered assets._

Cloudflare is our infrastructure provider. We run on Workers (EU region available) + R2 (S3-compatible object storage).

You don't configure this. We do.

### Supabase — auth + database
_Auth via Supabase OAuth flow; data in Supabase Postgres with RLS._

Supabase handles:

- Authentication (passwordless magic-link email sign-in + mandatory TOTP two-factor; invitation-only — no public self-signup)
- Database (Postgres with row-level security for tenant isolation)
- Vault (encrypted secrets storage)

EU region for EU tenants.


## Roadmap
Where Entradigm is going, in calendar-quarter granularity.

### Q3 2026 — BYO LLM routing + GDPR export jobs
_BYO AI provider becomes routable per-cycle; GDPR export becomes self-service._

Wave 5 (BYO LLM routing): strategist-worker adapter layer so each cycle picks the configured provider + model per tenant. Means daily cost moves to your provider account.

Wave 8 (GDPR export jobs): self-service Art. 15 / 17 / 20 endpoints; export jobs run in background, deliverable as signed R2 download.

### Q4 2026 — Lead-capture for tenants
_CTA-driven lead-capture flows + email aliases per platform._

Tier 3 work: per-tenant CTA keywords + per-platform email aliases. Inbound replies route through our processor into your CRM via webhook.

Optional Pro-tier billing layer (Stripe-based) lands here too.

### 2027 — Public agent API + MCP server
_The platform becomes formally agent-callable via MCP._

Today, the API exists internally + is documented. 2027 ships the formal public MCP server: a single endpoint where any LLM agent can wrap Entradigm as a tool with full schema introspection.

Means: an external agent (Devin, Cognition, your custom n8n flow) can trigger cycles, approve / park, edit brand rules, all programmatically.

### Long-term — agent-to-agent marketplace
_Tenants expose their content corpus to other agents under licensed terms._

Speculative direction: tenants can opt-in to expose their content corpus + brand rules to other agents under licensed terms. An agent serving a different tenant can subscribe to your industry coverage, your brand voice, your fact-checked content library.

Like a Spotify for industry-specific agent-callable content. No timeline yet.


## Support & help
When you need a human, here's how to reach us.

### In-app help — categorized FAQs
_The /app/help page has categorized FAQs + a contact form._

Most common questions are answered in the categorized FAQ at /app/help.

Categories: account · billing · brand setup · cycles · publishing · compliance · troubleshooting.

If your question isn't there, the contact form at the bottom of /app/help routes to the right human inbox.

### Direct contact channels
_Email + Telegram channel + (planned) community Discord._

For non-urgent questions: contact form at /app/help routes to operator inbox.

For urgent (cycle-blocking) issues: send a Telegram message to the BotFather bot you configured — the operator sees it within an hour during EU business hours.

Community Discord: planned post-Mon-2026-06-09.

### Platform status page
_Coming Q3 2026 — status.entradigm.com._

Live status page at status.entradigm.com will show Worker / database / Anthropic / Blotato availability + 90-day uptime + incident timeline.

Today: Cloudflare status page covers the infrastructure layer.

<<<ENTRADIGM-DOCS-END>>>

## Live schema endpoints

These PUBLIC GET endpoints return the JSON schema of every intent + the tool
registry. They are CORS-enabled (Access-Control-Allow-Origin: *) and require
no auth — introspect the contract before you call the bearer-gated POSTs.

- GET https://entradigm.com/strategist/schema
    Intent + response + slides + provenance schemas; apiVersion + capabilities.
    curl -s https://entradigm.com/strategist/schema

- GET https://entradigm.com/strategist/tools/schema
    The agent tool registry (5 tools).
    curl -s https://entradigm.com/strategist/tools/schema

- GET https://entradigm.com/memory/schema
    The check-memory (dedup) intent schema.
    curl -s https://entradigm.com/memory/schema

- GET https://entradigm.com/library/schema
    The ingest-library-file + query-library intent schemas.
    curl -s https://entradigm.com/library/schema