FAQ

General

What is AgenFleet?

AgenFleet is a platform for deploying and managing AI agents. You configure agents with a model, identity, tools, and instructions — AgenFleet handles the infrastructure, scheduling, monitoring, and cost controls.

How is AgenFleet different from calling an AI API directly?

A direct API call is stateless — no memory, no persistence, no scheduling. AgenFleet adds:

Persistent memory (agents remember context across sessions)
Cron scheduling (agents work autonomously on a schedule)
Tool access (agents can take action: search the web, call APIs, send messages)
Fleet monitoring (visibility into all agents’ health, cost, and activity)
Multi-tenant isolation (for organizations running agents for multiple clients)

AgenFleet works with your existing provider API keys — you bring your own credentials and token costs go directly to your provider account.

What models does AgenFleet support?

Currently the Claude model family (Haiku 4.5, Sonnet 4.5, Opus 4.6) via your Anthropic API key. Support for additional providers — including OpenAI, Google Gemini, Mistral, and others — is on the roadmap. See Models & Fallbacks for details.

What’s the difference between a session and a cron job?

A session is an interactive conversation thread — a human talking to an agent in real time. A cron job is a scheduled task that runs the agent automatically at a set time, without human involvement, and delivers output to a notification channel.

Agents

How many agents can I deploy?

Operational and Enterprise plans have no agent limit. Foundation plans are limited to 3 agents. See the pricing page for plan details.

Can two agents share memory or talk to each other?

Not directly. Agent memory is private and scoped per agent. To share information between agents, you can:

Use a webhook delivery on one agent’s cron job to trigger a second agent
Have one agent write to a shared workspace file that the other reads
Use n8n as an orchestration layer between agent outputs and inputs

Can I change an agent’s model after deploying it?

Yes. Go to the agent’s Settings tab → Configuration → update the model.default field and save. The change takes effect on the next session turn or cron job run. There is no downtime.

My agent’s responses are getting slower over time. Why?

Almost always session bloat. As sessions accumulate history, each turn sends more context to the model, which takes longer to process and costs more. Open the agent’s Sessions tab and archive or prune sessions that haven’t been used recently. Response speed will return to normal immediately.

Can I give an agent access to my internal systems?

Yes, via the http_request tool and environment variables for credentials. The agent can call your internal APIs, read data, and push updates. See Tools & Integrations for setup details.

Cron Jobs

What timezone do cron schedules run in?

UTC by default. Set the timezone field on the cron job if you need schedules in your local time (e.g., “every morning at 8 AM EST”).

What happens if a cron job fails?

AgenFleet retries once automatically. If the retry also fails, you receive an alert via your notification channel. The job is marked failed for that run, and the next scheduled run proceeds normally. Failed runs are logged in the agent’s Activity tab with the failure reason.

Can I run a cron job manually?

Yes. Open the cron job and click Run Now. This triggers an immediate execution outside the normal schedule, using the same prompt and delivery settings.

Can I send a cron job’s output to multiple channels simultaneously?

Yes. Add multiple delivery targets (e.g., Telegram + Slack + email) when configuring the job. All targets receive the same output.

Billing & Cost

How does billing work?

AgenFleet charges a monthly subscription for platform access — fleet infrastructure, scheduling, monitoring, and support. Token costs are not charged by AgenFleet. Instead, you connect your own AI provider API keys (e.g., Anthropic), and token usage is billed directly by your provider to your account at their standard rates.

See Settings → Billing for your AgenFleet subscription details, and Settings → Integrations to manage your provider API keys.

Why did my agent pause mid-day?

The agent reached its daily token budget. This is a safeguard to prevent runaway usage from bugs or loops — it does not affect your AgenFleet subscription. Configure limits.dailyTokenBudget in the agent’s settings to a higher value if legitimate usage is being cut off. The agent resumes automatically at midnight UTC when the budget resets.

How do I reduce costs without degrading agent quality?

The three highest-impact changes:

Switch routine cron-job agents from Sonnet to Haiku
Prune old sessions (reduces input tokens per turn)
Lower topK memory injection from 5 to 2–3 for agents that don’t need deep historical context

See Cost Management for a full optimization guide.

Security & Data

Is my data shared with other customers?

No. Each tenant’s data is isolated via row-level security policies enforced at the database layer. Even on shared infrastructure, one tenant’s data is structurally inaccessible to another tenant. See Data Isolation for the full architecture.

Can AgenFleet staff read my agent conversations?

AgenFleet staff can access session data for support and security purposes, subject to our privacy policy. We do not use customer conversation data to train AI models. For Enterprise plans, we can execute a DPA with stricter access controls.

How do I report a security vulnerability?

Email security@agenfleet.ai. We respond within 2 business days and do not pursue legal action against researchers who report issues responsibly.