DeskFerry

DeskFerry is a hosted no-code platform for building recurring AI agents that work across business applications. Users can describe a job in plain English, connect the relevant tools and run the agent on schedules or application triggers. The current product combines a simpler Assist surface with a multi-step Agent Builder and supports shared skills, knowledge, memory, human approvals and activity logs.¹

Its breadth comes largely from integrations: DeskFerry advertises more than 1,500 app connections through Composio, including Gmail, Slack, Notion, HubSpot, Salesforce, Google Sheets, Zendesk, Jira and Stripe, plus custom API connections. Agents can use models from OpenAI, Anthropic, Google, xAI and others without requiring users to maintain separate model API keys.¹²

DeskFerry offers a seven-day trial without a card, not a permanent free production tier. Starter costs $19/month for 2,000 credits and 400 actions; Growth is $49 for 5,000 credits and 1,000 actions; Pro is $99 for 10,000 credits and 2,000 actions. Every paid tier allows unlimited agents, while higher tiers add premium integrations, larger knowledge stores, bulk actions and team/workspace features.³

Model and third-party API consumption uses Vendor Credits. DeskFerry says those top-ups are charged at the provider's underlying API cost without markup and remain valid while the plan is active.³ This makes headline subscription price only one part of total cost for high-volume or model-intensive workflows.

The public product surface is unusually detailed about mechanics and commercial limits. DeskFerry documents schedules, triggers, approvals, knowledge, memory, run logs, app tiers, credits and action allowances rather than describing agents only in abstract terms.¹³ Its security material states that data and credentials are encrypted in transit and at rest, workspaces are isolated, and team access can be separated into admin, editor and viewer roles.⁴

The important limitation is evidence provenance. DeskFerry publishes customer testimonials and outcome figures, including lead-volume, response-time and hours-saved claims, but those examples are hosted by DeskFerry and are not independently audited.¹ Current external coverage does not provide a substantial verified-user rating corpus or independent comparative test of DeskFerry's execution reliability, connector failure handling or long-running agent accuracy.

DeskFerry is also fully hosted rather than a self-hosted framework.² Organizations with sensitive data should therefore evaluate its OAuth scopes, data retention, subprocessors and model-provider routing against their own policies rather than inferring privacy from the presence of encryption alone.

The strongest fit is a small or midsize team with repeated operational work spread across SaaS applications: lead qualification, support triage, reporting, invoice follow-up, CRM updates or content operations. DeskFerry removes much of the engineering work normally required to connect these systems and provides human approvals for actions that should not run entirely unattended.¹

It is a weaker fit for teams that require self-hosting, want code-level control of every orchestration primitive, or cannot tolerate usage-based credit/action limits. Organizations already deeply invested in a mature automation stack should also compare the incremental value of agent reasoning against the predictability of existing deterministic workflows.