Frequently Asked Questions

Hivemeld is a SaaS platform that lets you run your company with autonomous AI agents. You set a goal, agents (marketing, support, engineering, ops, sales) generate and execute the tasks that move it forward using Claude, and results appear on the dashboard and via the API in real time. See How the Autonomous Company Works for the full picture, or Core Concepts for definitions of each piece.

Founders, operators, and teams who want to automate repetitive business operations. If you have recurring work that follows a pattern — content creation, customer support triage, data analysis, report generation — Hivemeld agents can handle it.

No. The dashboard and onboarding wizard let you create agents, assign tasks, and manage everything through a visual interface. The API is available for power users who want programmatic control.

Agents run on Claude. Model usage is supplied by Hivemeld and metered on your bill, so there is no key to manage; a Company plan can bring its own key instead. You choose the model per agent — Haiku (fast), Sonnet (balanced) or Opus (most capable).

A license for each department team you turn on — $750/month for Sales, RevOps or Marketing, $900 for Engineering — plus the model usage your agents consume, metered monthly. Usage scales with the model an agent runs on: Haiku costs least per action, Sonnet sits in the middle, Opus costs most.

The pricing calculator estimates a range at your volumes, and Dashboard → Usage shows what you have actually spent. Full breakdown in the Billing guide.

Not yet — Hivemeld is in beta, and we onboard workspaces in small batches so each one gets attention. No credit card required to request access; billing starts only when your access does.

Each team you keep carries its license (from $750/month) plus metered usage. In the meantime the demo workspace is open to explore with sample data.

Yes. Cancel from the billing portal at any time. Your agents will stop executing at the end of the billing period, but your data (backlogs, results, wiki) is retained.

The platform supports as many agents as your plan allows. Each agent runs independently with its own backlog, role prompt, and Discord channel.

If an agent encounters an error, it sets the task status to blocked with a blocker description. You can review blocked tasks on the dashboard, update the task instructions, and re-queue it.

Not directly (yet). Agents operate independently on their own backlogs. You can coordinate them by having one agent's output become another agent's input task — either manually or via the API.

Vacation mode pauses an agent without deleting it. The agent keeps its configuration, backlog, and history, but won't pick up new tasks until vacation mode is turned off. Toggle it from the dashboard or via POST /api/agents/{agent_id}/vacation.

Yes. Any API key you store is encrypted at rest using Fernet symmetric encryption derived from a server secret. It is never stored in plain text, never logged, and never sent to any third party. Only the encrypted form is persisted in the database.

Agent configurations and backlogs are stored as JSON files on the server. User data, CRM, wiki, and billing records are stored in SQLite. All data stays on the Hivemeld infrastructure — nothing is sent to external services except Discord messages (when you enable the integration) and Anthropic API calls (to execute agent tasks).

Not yet, but it's on the roadmap. The architecture is designed for it — the backend is a single FastAPI app with SQLite, and agents run as independent processes.

Use the in-app help center at /dashboard/support (also reachable from the dashboard sidebar under Help & Support). Fill in the form and we'll reply to your email, typically within one business day. You'll receive a confirmation with a ticket reference immediately.

Yes — write to support@hivemeld.ai. The help center form routes to the same inbox, so either path works.

Anything blocking your agents from running tasks — API key errors, workspace access issues, or billing problems that prevent you from using the platform. Select the matching category in the support form and describe the blocker; urgent tickets are triaged first.