CogniLead, documented.

The content plane turns a product description, a topic, and a keyword set into draft content. It generates long-form blog-page drafts, conversion-oriented landing-page drafts, and structured on-page recommendations (titles, meta descriptions, heading outlines, link suggestions). Everything is a single structured LLM call to phi-cloud — useful scaffolding to start from and edit, not keyword research, SERP analysis, or schema markup.

What it produces

• Blog-page drafts — long-form articles targeting an informational keyword cluster.
• Landing-page drafts — conversion-oriented copy targeting a commercial-intent keyword.
• On-page recommendations — title tag, meta description, H1/H2 outline, and link suggestions, returned as structured fields. These are LLM suggestions, not a computed link graph.

How to call it

lib/seo/generate.ts is the in-process entry point. Over HTTP, POST /api/v1/seo/generate takes the product, topic, keyword set, and content kind, and returns the generated artifact plus the on-page recommendations. The same capability is exposed as a free, capped MCP tool for AI IDEs (see §6).

POST /api/v1/seo/generate
{
  "product": "string",
  "topic": "string",
  "keywords": ["string", "string"],
  "kind": "blog" | "landing"
}

The outbound email plane sends personalised, multi-step sequences to the leads you supply via POST /api/v1/leads. Dispatch is intentionally thin — Resend handles the SMTP, the bounce tracking, and the reply detection. CogniLead owns the warmed sender-domain pool, personalisation (via phi-cloud), the sequence schedule, and the outbound hygiene that keeps the channel healthy.

Delivery

Each sequence step is rendered and handed to Resend for delivery. Inbound webhooks stream bounces, replies, and complaints back into the tenant's data, which drive the suppression list and the per-step state of the sequence.

Personalisation

Drafts are generated per lead via phi-cloud, grounded in the lead data you supply. The output is structured (subject, body, html, text) so it can be validated before it is queued for delivery.

Hygiene

CogniLead keeps only the outbound hygiene the channel actually needs: a per-tenant suppression list and an RFC 8058 one-click unsubscribe. A complaint, a bounce, or an unsubscribe against one campaign suppresses the address across all of that tenant's campaigns. That is the whole compliance surface — see §5.

CogniLead does not ingest, enrich, or rank a company universe. It is the deliverability layer downstream of your lead source (your CRM, a scraper, an Apollo or Clay export). You hand it leads; it gets the mail into the inbox.

Deliverability keeps the outbound channel healthy and the compliance surface honest. CogniLead leans on Resend for the heavy lifting — SMTP, bounce handling, reply detection — and keeps its own footprint to the two hygiene primitives outbound actually requires.

Suppression list

Each tenant has a global suppression list. An email or domain on the list is a hard gate — it is checked before any send, and a hit drops the lead. Bounces, complaints, manual entries, and unsubscribes all write to the same list, so a suppression against one campaign suppresses the address across every campaign that tenant runs.

One-click unsubscribe

Every message carries an RFC 8058 one-click unsubscribe in List-Unsubscribe-Post. An unsubscribe is honoured by adding the address to the suppression list immediately, across all of the tenant's pools.

The MCP server is the acquisition lever. A developer running an MCP-aware IDE can install @cognilead/mcp-server and reach the free content tools without a key, then authenticate to unlock the gated outbound tools — without leaving the editor.

Tools

• cognilead.draft_pitch — free, capped (20/day). Drafts an email pitch from the context you supply.
• cognilead.score_signal — free, capped (50/day). Scores a lead or signal you supply for product fit (0–100, deterministic).
• cognilead.ingest_lead — gated. Registers a customer-supplied lead in the pipeline (suppression + dedup on entry).
• cognilead.run_campaign — gated. Kicks off an outbound campaign by id with an optional lead filter.
• cognilead.suppress — gated. Adds an email or domain to the suppression list.

A gated cognilead.generate_seo_content tool is also exposed for AI content scaffolding (see §2). Targeting is customer-supplied — you bring the leads (see §4).

Why MCP at all

Developers discover CogniLead inside their IDE, not from an ad. The free tools cost only the phi-cloud tokens they consume, and the gated tools return a 401 with an inline signup URL so the conversion path is built into the call. See /mcp for the install command.

The API surface is the same set of capabilities the dashboard and the MCP server expose, reachable over HTTP. The contract is published early so tenants can build against it; the official TypeScript and Python SDKs follow.

Endpoints

POST   /api/v1/leads
GET    /api/v1/leads
GET    /api/v1/sends
POST   /api/v1/seo/generate
GET    /api/v1/suppressions
POST   /api/v1/suppressions

Authentication

Per-tenant pk_live_* (publishable) and sk_live_* (secret) keys. Argon2id hash at rest. Last-4 displayed only at creation. Both ride in Authorization: Bearer <key>. Rotation reminders fire at 90 days; secret keys are revocable on a 60-second propagation window.

SEO generation

POST /api/v1/seo/generate is the workhorse endpoint — it backs lib/seo/generate.ts and returns the generated blog or landing page plus the on-page recommendations. It is the one endpoint also exposed free (and capped) through the MCP server.