Freshdesk AI agent: Freddy features, pricing, and eesel CLI

What is the Freshdesk AI agent?

Freddy AI Agent is the customer-facing part of Freshworks' AI offering. It can answer from configured knowledge and use configured workflows. AI Copilot and AI Insights do different jobs; buying an assistance feature is not the same as deploying an agent to customers.

The Freshdesk pricing page separates these features. Avoid treating every item branded "Freddy AI" as one bundled purchase.

There is also a product distinction. Freshdesk lists its Email AI Agent, while Freshdesk Omni includes conversational and email agents. Confirm your specific product, plan, and channel before following setup instructions or comparing prices.

What Freddy AI Agent can do

Freshworks' AI Agent Studio guide describes knowledge, workflows, instructions, testing, deployment, and monitoring. In its customer-support setup, the documented knowledge inputs include public URLs, supported files, solution articles, and custom Q&As.

Freshworks' knowledge configuration guide shows these source types and their learning status.

The workflow builder and library extend what the agent can do beyond giving an answer. Evaluate the exact operation you need, including its inputs and failure path. A connection to a payment system does not by itself prove that the intended refund operation is configured or authorized.

Instructions shape conversation behavior and business rules. They do not eliminate the need to test unsupported questions, conflicting information, and requests for a human. Do not read a broad resolution-rate marketing claim as a prediction for your queue.

For more detail on source coverage, see the Freddy knowledge guide. The important check is whether the actual content needed for your support cases is available, current, and appropriate for that audience.

How to set up the native Freshdesk AI agent

In AI Agent Studio, an admin creates an agent, adds knowledge, configures workflows and instructions, tests it, and maps it to the intended channel. The Studio guide covers Freshdesk Omni and Freshchat; use the setup instructions for your actual Freshdesk product rather than assuming channel parity.

Freshworks provides a shareable preview for review. Its documented preview does not perform agent transfers for unanswered queries or requests for human help. Verify that behavior in the deployed channel.

A practical rollout checklist goes beyond whether the first answer sounds right:

• Check one question that the approved knowledge answers directly.
• Ask a question where an essential detail is missing.
• Test a policy exception and a request to speak with a person.
• Inspect the actual action result, not just a message claiming success.
• Confirm what existing routing and automation rules do to the same conversation.

Keep the initial scope narrow enough that a human can inspect every result. If a test fails, identify whether the cause is missing knowledge, the instructions, permissions, or the delivery workflow before changing several things at once.

Freshdesk AI agent pricing

The base prices below were checked on September 8, 2026. They are per human agent per month, billed annually, before separate usage and add-ons.

Use the official Freshdesk plan table for email-focused service, or the Omni plan table for that product.

Additional Freddy AI Agent usage is $49 per 100 sessions after a one-time 500-session allowance per account, available on Growth, Pro, and Enterprise. The current email-session definition is a 72-hour window from the customer's first email. Omni's chat-session window is 24 hours. Pack validity follows the payment cycle.

For 2,000 sessions, an unused 500-session allowance leaves 1,500 paid sessions: 15 packs × $49 = $735. Once the allowance is gone, 20 packs cost $980, before helpdesk licenses. This is illustrative pack arithmetic, not a quote for every contract.

Usage pricing does not necessarily mean better automation is penalized. You need to compare the bill with the work completed, escalations, and human correction required. A low per-session price is not helpful if the answers create rework; a higher bill can still be worthwhile if it reliably removes work your team would otherwise do.

Where an eesel teammate fits in Freshdesk

eesel's Freshdesk integration lets a support teammate work in the queue you already use. It can draw on approved knowledge, draft private notes, or perform enabled ticket actions. The CLI adds another way to configure and inspect that teammate without asking every technical colleague to work through dashboard screens.

The integration documentation separates two connection modes:

Quick start reads the public help center using your Freshdesk subdomain. It does not provide tickets, canned responses, triggers, or actions.

Full connection uses the subdomain and a Freshdesk API key. Enter the key in eesel's secure form, not in a chat message or shell command. It enables access to the published help center, canned responses, and resolved or closed past tickets. Review source toggles and optional PII redaction, then check sync status.

Connected content is not automatically safe for every audience. Historical tickets can contain outdated advice or internal information. Decide which sources the teammate should use and test whether customer-facing answers expose anything they should not.

Connecting alone does not create a new ticket automation. However, an existing enabled automation may already be running, so inspect the agent's configuration before treating it as a test environment.

Prepare a Freshdesk pilot through eesel CLI

The useful CLI workflow here is to separate three checks: what the teammate knows, what it is allowed to do, and what actually happened on the ticket. A technical teammate or coding agent can help with each, while support staff review the same agent in the dashboard.

Inspect the selected agent

For this example, use an existing eesel workspace and select or create a separate test agent in the dashboard. Replace TEST_AGENT_ID with that agent's ID. Connect only an account and knowledge you are authorized to use for testing.

With Node.js 18.17 or newer:

npx @eesel/cli login
npx @eesel/cli whoami
npx @eesel/cli agents
npx @eesel/cli --agent TEST_AGENT_ID status
npx @eesel/cli --agent TEST_AGENT_ID integrations
npx @eesel/cli --agent TEST_AGENT_ID instructions
npx @eesel/cli --agent TEST_AGENT_ID automations

Check the signed-in workspace, available source counts, standing instructions, and any enabled automations. A separate agent does not isolate Freshdesk data or actions if you connect it to a live account.

For a new connection, the CLI also supports integrations connect freshdesk. Follow the secure connection flow and check the current CLI reference before using additional options. An API key does not belong in an example prompt or a committed configuration file.

Test a reply before enabling writes

Start with a fictional question and an explicit expected result. In the example below, the teammate should ask for diagnostic details rather than claim to have fixed the customer's export.

npx @eesel/cli --agent TEST_AGENT_ID new --name "freshdesk-export-pilot"
npx @eesel/cli --agent TEST_AGENT_ID chat "Fictional support test. Policy: for a report-export timeout, ask for the report date range and file format. Do not claim to retry the export or change account settings. Customer: My monthly report will not export. Draft a reply here for a reviewer, and identify the missing details. Do not post to Freshdesk or change anything."

This is an illustrative command sequence, not a test I ran against a customer's account. Before running it, keep write actions off or require approval in the agent's actual settings. The "do not post" instruction is not an access-control mechanism.

The policy above is test context, not a persistent instruction update. After checking this simple case, evaluate your approved real knowledge with additional cases that were not used while editing the instructions. Include a customer who already provided the requested details, so the agent should not ask again.

The new command starts a conversation, not another agent. Ordinary chat continues the selected conversation. Chat can invoke tools and run billable work, so inspect permissions first and keep early evaluations small.

Configure and verify the private-note workflow

Confirm that your Freshdesk plan supports the webhooks required for ticket triggers. eesel's integration documentation lists Sprout/free as unsupported for these triggers, so the ticket pilot below will not run on that plan.

Once the answer tests pass, choose a limited ticket trigger and filter in the dashboard or ask the teammate to configure them in chat. Review the proposed scope before authorizing the change. Creating an automation can enable it immediately; do not create an unrestricted new-ticket rule on a live account just to see what happens.

For an initial pilot, use Leave internal note and keep Send ticket reply off. The first leaves a draft for your team; the second sends to the customer. Actions can be off, approval-required, or automatic. Keep unrelated writes disabled, especially ticket merging, which Freshdesk cannot undo.

Trigger the pilot with a test ticket you own or are authorized to modify, then inspect:

npx @eesel/cli --agent TEST_AGENT_ID activity
npx @eesel/cli --agent TEST_AGENT_ID approvals
npx @eesel/cli billing

Do not count a held approval as a completed action. Similarly, a good terminal answer does not prove the ticket trigger or note delivery works.

Give a coding agent a bounded review task

Claude Code, Cursor, or Codex can run the inspection commands and read the results. Single results are JSON; lists contain one object per line. That makes the setup review useful for scripts and coding agents, not just a human typing in a terminal.

For example, ask:

Inspect eesel test agent TEST_AGENT_ID through the CLI. Summarize connected Freshdesk sources, existing automations, and the export-test result. Flag missing knowledge or held approvals. Do not change instructions, permissions, or automations, and do not approve actions.

The coding agent is operating your eesel teammate, not Freddy. Its own subscription does not cover eesel usage. Also, CLI --dry-run previews a server write call; it does not simulate support quality or calculate a future resolution rate.

Compare costs and choose the next step

eesel's current pricing bills a regular support ticket or chat-session task at $0.40, regardless of outcome. Multiple replies within the same ticket do not become separate message charges. Default pay-as-you-go has no platform fee or monthly minimum; Enterprise adds a platform fee.

That is not "pay only when resolved." A spending cap can pause new work, but it does not make a usage-based bill fixed. Compare costs using the same support workload and include the time your team spends reviewing or correcting answers.

If you are already considering a helpdesk migration, evaluate the destination platform's native agent as part of that separate decision. A tool for another helpdesk is not automatically a drop-in Freshdesk alternative. For this article's stay-on-Freshdesk use case, compare Freddy with an eesel pilot in your existing workflow.

This dashboard example shows a Zendesk-connected teammate. The same eesel workspace is accessible through the CLI; your Freshdesk pilot should be checked in its own connected account.

Try eesel with one Freshdesk support topic and a private-note workflow. Use the CLI to inspect setup and evidence, then widen the scope only after the actual ticket behavior meets your review criteria.