Add AI agents to your helpdesk, Slack and store. Draft replies, route tickets and perform approved actions using your company's knowledge.
Used this tool? Rate it
Used this tool? Rate it
eesel (also branded eesel AI) provides AI agents for customer service, content and operations inside the apps a team already uses. Its AI customer support agent uses company knowledge to prepare replies and perform permitted work in an existing helpdesk. Content and ecommerce agents extend the service beyond support.
The helpdesk agent can draft replies for review or send them itself, triage, tag and route tickets, and answer from a help center, macros and past tickets. It can assist a person with drafts or run enabled actions automatically. Autonomy is vendor-described, not a measured resolution rate.
Permissions govern available actions, instructions govern behaviour, and automations govern events. Instructions cannot override an approval requirement.
Slack users can mention the agent in a channel or send a direct message. It answers in the thread from connected knowledge, links the source and handles follow-ups without another mention. It also looks up records and drafts replies in connected apps from that thread, separately from Slack's built-in AI.
The Shopify AI agent answers shoppers in website chat, shows product cards with Add to Cart, and looks up live stock, orders and tracking. Its sources include products, pages, blog content and policies; actions include returns and shipping-address changes.
Show Cart displays the live cart, while checkout remains in Shopify checkout.
Prices were captured on October 29, 2026. The $299 starting price is identified as USD in structured pricing data. Other amounts are displayed with a dollar symbol.
The trial provides 100 credits for seven days, with no card and every feature unlocked. It ends at either limit and agents pause. Up to two blog posts and 20 images can be generated, consuming the same trial credits.
| Monthly credits | Displayed monthly price |
|---|---|
| 500 | $299 |
| 1,000 | $499 |
| 1,500 | $699 |
| 2,000 | $849 |
| 2,500 | $999 |
| 3,000 | $1,149 |
| 3,500 | $1,299 |
| 4,000 | $1,449 |
| 4,500 | $1,599 |
| 5,000 | $1,749 |
| More than 5,000 | Contact the team |
Teammate includes every integration and channel, scheduled agents, reports, skills, unlimited agents and seats, and community support. Channels include helpdesk, Slack, chat bubble, email and extension.
Enterprise has custom pricing and adds a dedicated setup contact, SSO and MFA, custom data residency, HIPAA with a signed BAA, security review, DPA and CSA, plus an annual contract and invoicing. Add-ons for any plan start at $299/month for priority support and $199/month for SAML SSO with enforced two-factor authentication. The annual upfront offer is two months free, for any batch size, arranged with the team; the captured page does not display separate annual amounts.
A ticket the agent works on normally uses one credit regardless of reply count. A chat-bubble, Slack or email conversation likewise uses one credit for the conversation. Scheduled runs cost at least one; generated images cost one each; a blog post costs several, including its images. Most dashboard chats are free, but a very long chat can use credits beyond its free allowance. Setup work the agent initiates itself is described as free.
Tasks consume credits regardless of outcome, according to the billing documentation. Support, content and commerce agents share the workspace's credits.
The marketing pricing FAQ says a ticket is one credit however many replies it takes and whatever it does inside that ticket, including order lookups or updates to another tool. The billing documentation says unusually heavy work, including a ticket requiring extensive research, can use more than one credit. The two descriptions set different ticket-cost boundaries.
The marketing pricing FAQ says agents pause when credits run out unless overage is turned on with a chosen monthly cap, at $0.80 per extra credit on the next invoice. The billing documentation instead describes agents continuing on $0.80 extra credits with an initial $200 cap; setting that cap to $0 makes them pause as soon as plan credits are used. Extras are invoiced at each $100 accumulated, plus the balance at month end. Default overage and invoice timing therefore differ.
Credits reset each billing month and unused credits do not roll over. At the extra-spend cap, work already in progress finishes while new work stops. Workspace settings > Billing shows the plan, usage, extra-spend cap and invoices. An upgrade takes effect immediately, starts a new billing month with full new-plan credits and credits the unused part of the old plan; a downgrade applies at the next billing date. Cancellation is available at any time, with access continuing to the end of the current billing period. Subscriptions automatically renew unless cancelled before renewal. For an SLA missed in two consecutive months or over three months in any 12-month period, the terms specify termination upon notice and a prorated refund of prepaid fees for the remaining billing period as the only remedy.
These are vendor descriptions, without independent performance comparison.
The terms do not guarantee that the service will always be safe, secure or error-free, or operate without disruptions, delays or imperfections.
The company says it does not train or fine-tune models on customer data, and that AI providers process it to generate responses under API terms prohibiting training. It describes workspace data as isolated, knowledge search as limited to supplied content, and tools as limited to those enabled. Outbound requests are restricted to allowed domains. App credentials go through a secure form to its servers; the company says the AI model does not see the passwords or API keys.
Residency descriptions differ in scope. The pricing cards list EU or US data residency among the included features. The Trust Center describes production in AWS Northern Virginia, with full EU data residency available on Enterprise. It separately describes EU workspaces as stored in AWS Ireland with an EU search index. Included-card wording and the Enterprise condition remain distinct public statements.
The optional Bedrock configuration routes model calls through eesel's own AWS account and can run in the US or EU. Tools that would send content to other services, such as web search, are disabled in that configuration. AWS remains a subprocessor and the agents still run in Vercel-hosted sandboxes; this is not a move of the entire service to customer-owned infrastructure.
SOC 2 Type II was in progress at the capture date. The company gives an observation period of October 1–December 31, 2026, followed by an independent audit in early 2027 and a report under NDA once issued.
Home-page marketing also says helpdesk access introduces no new data pipeline or third-party storage. Other documentation explicitly describes stored ticket history and a search index; these are different storage descriptions.
PII redaction applies to past helpdesk tickets synced into persistent knowledge. The toggle currently appears on Zendesk ticket settings; other helpdesks use the workspace default without their own switch. Recent workspaces default to redaction off, older workspaces default to on, and an explicit saved setting takes precedence.
Enabling redaction does not retroactively redact already-indexed tickets; a fresh past-ticket sync is needed for older material. Runtime tool calls are unaffected: a live Zendesk ticket fetched for drafting reaches the model as raw content. Articles, macros, uploaded files and other non-helpdesk knowledge sources are not covered by that redaction path.
Deletion has different published scopes and timing. The Trust Center says deleting a workspace immediately wipes live documents, search index, integrations, credentials, agents and conversation history; encrypted backups roll off within 35 days and security logs expire on a fixed schedule. The privacy policy says Customer Content is deleted within 60 days of a request after termination, with some information retained for legal obligations, disputes or agreement enforcement. These are different deletion scopes, not immediate removal of every copy.
The company says the helpdesk agent supports more than 80 languages and answers in the customer's language using multilingual ticket history. This is a vendor statement, not a per-language quality measurement.
The documented sync fails closed: tickets that cannot be redacted are skipped rather than indexed raw, then retried on the next sync. For redaction, the company says the redacted version replaces the original before database persistence or search indexing. The live-fetch exceptions still apply.
Only the Owner can delete the workspace, and a paid plan must be cancelled first; otherwise Delete workspace is disabled. Deletion is permanent and removes agents, settings, connected apps, knowledge and past chats. There is no separate account-deletion path. Plans purchased through Atlassian Marketplace must be cancelled there.
The terms describe the trial as genuine evaluation, once per individual, workspace and organization, and prohibit production workloads, resale and bulk-output extraction without intent to subscribe. The standard seven-day trial may be extended once for another seven days, up to 14 in total, subject to the provider's communicated usage limits.
