Every feature, built for delivery work.
Helm has six modules. Each one solves a specific problem an agency has. Nothing for the sake of it. No separate tiers — every workspace gets the whole product.
Tasks built for delivery work.
Most task tools optimise for engineering sprints. Helm optimises for the lifecycle of an agency deliverable — TODO → IN_PROGRESS → READY_FOR_REVIEW → WITH_CLIENT → DONE.
- Status flow that mirrors how delivery actually moves (with rework + awaiting-input states)
- Priority, due date, time estimate, assignee, collaborators, recurring templates
- Workload view: each operator's committed vs free hours, by week
- Per-client filter, search, custom sort — every task page is one click from any view
- Time entries log against tasks; client-portal time per task is auto-aggregated
Every client in full context.
A client page that's actually useful — not just a list of tasks. AI-built profile, recent feedback, hour-cap progress, stakeholder list, and the linked credential vault.
- Auto-generated client profile (cadence, preferences, pain points, recent context)
- Plan + monthly hour cap with warning thresholds
- Stakeholder list with optional portal logins
- Time sheets grouped by week, billable vs non-billable
- Activity log of every change, search-able and filterable
Time tracking that doesn't feel like work.
Start a timer with one click on any task. Stop it. Done. Logged against the task, the client, and the operator. No filling spreadsheets at 5pm Friday.
- One-click timers (only one running per operator at a time)
- Manual entries when you forgot to start the timer
- Billable / non-billable per entry; client-portal shows billable only
- Per-client monthly cap — alerts when 80% / 100% reached
- Workload view: committed (estimates) vs logged (actuals) per operator
A credential vault, not a Google Doc.
Per-client encrypted vault. AES-256-GCM, secrets server-side only, every reveal logged. The client portal sees only the credentials you flag as shareable.
- AES-256-GCM with a per-deployment encryption key
- Per-credential "share with client" toggle
- Audit log of every reveal (who, when)
- Search by platform / username
- Notes field for context (not the secret itself)
An inbox that turns email into tasks.
Connect Gmail. Helm watches the inbox, parses incoming emails with AI, and creates review-ready task drafts. You approve, edit, or reject — Helm never auto-creates real tasks.
- Gmail OAuth (read-only)
- AI parses subject + body into draft task / comment / credential / feedback
- Per-item approve / reject flow — bulk-approve from inbox
- Confidence score + reasoning for every parsed item
- Noise filter (LinkedIn alerts, OTPs, newsletters auto-marked as IGNORED)
A client portal that doesn't suck.
One link, the client's brand, the work that's been done, the hours billed, invoices to pay, credentials to copy. No login required — magic-link or shared access code.
- Per-client branding (logo, primary color)
- Live mode (every visible task updates in real-time) or Weekly mode (admin-curated digest)
- Invoices visible + payable from the portal
- Time spent per task, aggregated for parallel work
- View-tracking: see when the client opened, viewed an invoice, clicked pay
AI you can turn off.
Every AI feature in Helm is part of the Starter plan ($100/mo). They run through one wrapper that checks a master kill switch and a per-org daily budget — flip the switch in admin settings and every Anthropic call short-circuits. Your tasks, clients, time tracking, and billing keep working — only AI is suspended.
Trial workspaces don't get AI by default — start your trial to try the rest of Helm first, then unlock AI when you subscribe.
Paste a WhatsApp message, meeting note, or chat log. Helm splits it into draft tasks, comments, credentials, and feedback — each routed to the right client.
Each Gmail message becomes a review-ready draft. You approve. No surprise auto-tasks, no AI hallucinating a deadline.
Cadence, preferences, pain points, recent context — Helm reads your tasks + feedback + emails and writes a markdown profile per client.
When a task moves to DONE (or you click "Suggest next"), Helm proposes 1–3 logical next-step tasks for that client.
Today's priorities, bottlenecks, at-risk clients, quick wins. One short Markdown brief generated overnight.
LinkedIn alerts, OTPs, newsletters, password resets — Helm marks them IGNORED so they don't clog your review queue.
Make it look like yours.
Set your logo and brand color in Settings → Workspace and the dashboard wears it. Per-client portals can carry the client's own brand — they never see a generic Helm screen.
- Org-level: logo + accent color in the dashboard sidebar
- Per-client: full portal theming (logo, primary, accent)
- Pasted URLs (no upload infra to deal with)
- Reset to Helm defaults any time
Changes apply across every page in the dashboard for everyone in your org.
Security & data
Every row in the database carries an organizationId. Queries are scoped at the auth layer — different workspaces never see each other's data.
AES-256-GCM with a 32-byte key held only on the server. Even with database access, the cipher text is useless without the key.
Passwords are bcrypt-hashed (10 rounds). Sessions are JWT-signed with a server-side secret. Demo-mode click-a-name is off in production.
Every Anthropic call is gated by a per-org kill switch and per-org daily budget. Logs are kept for 30 days for the admin health widget.
Helm runs on managed Postgres (Neon) with continuous backups. We don't train AI on your data, ever. Export to CSV any time.
Every server action verifies the actor's organization before touching a row. Compound foreign-key uniqueness (orgId + slug) prevents leaks.
Now do the part where you try it.
7-day trial. No card.
Start your trial