Skip to main content

01

Prompt from the LLM you already use

In Claude Cowork, ChatGPT, Cursor, or n8n, ask the agent to follow up, schedule, triage, or pull context.

02

Helix supplies the identity layer

The LLM calls Helix over MCP or REST. Helix reads the right inbox, calendar, contacts, and policy for that human.

03

Approve the real-world action

Drafts, sends, and calendar changes appear in Helix with source context, boundaries, and a full audit trail.

Claude, ChatGPT, Cursor, n8n

Helix does not replace the LLM. It gives the LLM a safe, human-backed way to use email and calendar.

01 · Human prompt

Ask Claude Cowork or ChatGPT

“Reply to Alex, move the launch check-in, and keep the tone direct.”

02 · Tool call

The harness calls Helix

search_mail → draft_reply → propose_calendar_update, all under that user’s identity.

03 · Human control

Helix pauses before impact

You see the draft, source thread, calendar diff, policy boundary, and Approve / Reject.

04 · Result returns

The LLM gets a receipt

Claude or ChatGPT can continue with proof: sent, scheduled, denied, or waiting.

Proven workflows

Start from a job, not a blank prompt.

Use templates for sales follow-ups, hiring coordination, and inbox triage — each with instructions, boundaries, and approval policy already wired.

Why teams trust it

Professional guardrails, visible by default.

Gmail + Outlook

Production connectors from day one

Every action logged

A clear audit trail for every draft and send

No surprise credits

A flat free tier while teams validate workflows

Self-healing tokens

Refresh and continuity handled behind the scenes

Secure your domain

Claim your agent’s domain.

Each *.nylas.email domain is permanent. Type the domain you want — we’ll tell you live if it’s open.

.nylas.email
Claim

Letters, numbers, and dashes. 3+ characters.

Newsletter

Get the Helix changelog.

Two emails a month. Shipped features, no marketing fluff.

No spam. Unsubscribe in one click.