Manage your digital workload with AI agents
Tell your pal the job. It searches, opens pages, clicks through sites and hands back a finished answer.
Reads public pages only. Unused credit goes back to your balance. Rules
The pages your pal opens in this task show up here, one step at a time.
Click any step to see what the Agent did there. Hide this pane for a plain chat.
Manage your digital workload with AI
Give Exepal Agent a research task. It searches and reads public web pages, then hands back a sourced answer you can copy or export. Pay per run in credits, from the website or by API.
Buy credits
Credits come from your Exepal balance, the same one every Exepal tool uses. 1 credit = $0.01. Credit never expires, and unused packs can be refunded within 14 days (refund policy).
How it works
1. You give it a task
Research, comparisons, summaries of public pages, pulling prices or specs into a table. Type your own or start from a ready-made task. You see an estimated cost before it starts and set the most it may spend.
2. It searches and reads
The agent searches the web, reads pages with the safe page reader (in a real browser tab when a page needs JavaScript), clicks through a site when the answer sits behind a tab, a pricing toggle or the site's own search, crawls a few pages of a site and pulls fields into tables. When the task needs it, it also runs Exepal's own checks: launch scan, competitor report, Name Check and SEO data, at their list price (the launch scan and AI readiness check are free). You watch each step live.
3. You get a sourced answer
Copy it, share it as a read-only link, save it as PDF, or download it as CSV, Markdown or JSON. Ask a follow-up to refine it, or press Watch this for me to get it by email every day or every week with what changed at the top. Put related tasks in a project so the agent follows your instructions, knows your notes and files, and remembers earlier answers. You pay only for what the run used, never more than your limit.
Pricing
10 free runs every day. Sign in and you get 10 free runs a day on DeepSeek Flash, each worth up to 5 credits, with no card needed. Your very first run is free on any model, up to 10 credits. The launch scan and AI readiness check cost nothing inside a run.
After that you pay only for what a run uses: the AI model's tokens plus each tool call. 1 credit = $0.01. Failed tool calls cost nothing, a run never spends more than the limit you set, and anything it doesn't use goes back to your balance.
What a real run costs: reading two pages and answering on DeepSeek Flash is about 1.7 credits ($0.02). A research task with two searches and five page reads is about 3 credits ($0.03) on Claude Haiku 5.5.
| Item | Credits | US dollars |
|---|---|---|
| DeepSeek V4.1 Flash model, per 1 million tokens (input / cached / output) | 5.7 / 5.7 / 75 | $0.057 / $0.057 / $0.75 |
| Pro model, per 1 million tokens (input / cached / output) | 180 / 18 / 300 | $1.80 / $0.18 / $3.00 |
| Claude Haiku 5.5 model, per 1 million tokens (input / cached / output) | 15 / 1.5 / 75 | $0.15 / $0.015 / $0.75 |
| Claude Sonnet 5.5 model, per 1 million tokens (input / cached / output) | 300 / 15 / 1500 | $3.00 / $0.15 / $15.00 |
| Claude Opus 5.5 model, per 1 million tokens (input / cached / output) | 600 / 30 / 3000 | $6.00 / $0.30 / $30.00 |
| Web search | 0.3 | $0.003 |
| Read a page | 0.15 | $0.0015 |
| Read a page in a real browser tab (JavaScript pages) | 0.75 | $0.0075 |
| Click through a site in a browser tab (tabs, toggles, search and filter boxes) | 0.75 | $0.0075 |
| Crawl a site (up to 8 pages) | 0.75 | $0.0075 |
| Extract fields from a page (plus model tokens) | 0.3 | $0.003 |
| Launch scan (Exepal check) | Free | Free |
| Competitor report (Exepal check, list price) | 50 | $0.50 |
| SEO report (Exepal check, list price) | 50 | $0.50 |
| Name Check full report (Exepal check, list price) | 25 | $0.25 |
Watched tasks (daily or weekly) cost nothing to set up: each repeat is paid like any other run and never spends more than the limit you pick.
What the agent will not do
In its own cloud browser it only works on public pages. In that browser tab it clicks and searches like a visitor, but it never signs in or signs up, pays, posts, sends messages or sends forms. With My Chrome (the Exepal Agent Chrome extension) it can also work in your own Chrome on sites you allow, where you are signed in; there it shows you anything it would send, save, post or delete before it happens, and it still never signs in, types passwords or card details, or pays (what it reads). In a research run it won't collect people's email addresses, phone numbers or home addresses, build lead lists, help with spam or mass messaging, create fake reviews or followers, or get past logins, paywalls, CAPTCHAs or robots.txt. Tasks like these are refused before anything is charged. Full rules: Terms, Agent section. Finding the public contact email of a business and writing to it is a separate product with its own limits: Email Outreach (Terms, Outreach section).
Questions
What is the Exepal Agent?
The Exepal Agent is an AI agent for web research. You type a task in your own words; it searches the web, reads and clicks through public pages, runs Exepal's own website checks when the task needs them, and hands back one answer with its sources.
How much does the Exepal Agent cost?
10 runs a day are free for every signed-in account on the DeepSeek Flash model, each worth up to 5 credits, and your first run is free on any model. After that you pay for what a run uses: 1 credit is $0.01, a typical research run costs about $0.02 to $0.03, and a run never spends more than the limit you set.
Which AI models can I choose?
DeepSeek V4.1 Flash, Claude Haiku 5.5, Claude Sonnet 5.5 and Claude Opus 5.5. The price of each model per million tokens is in the table on this page.
Can it repeat a task every day or every week?
Yes. Press Watch this for me on a finished task and it runs again daily or weekly, with the answer sent by email and what changed at the top. Each repeat is paid like a normal run and never spends more than your limit.
Can it work in my own browser?
Yes, with My Chrome, the Exepal Agent Chrome extension. It works only on sites you allow, shows you anything it would send, save, post or delete before it happens, and never signs in, types passwords or card details, or pays.
What will the agent not do?
In a research run it will not collect people's personal contact details, build lead lists, help with spam, write fake reviews or get past logins, paywalls, CAPTCHAs or robots.txt. Tasks like these are refused before anything is charged.
Can I call the agent from my own code?
Yes. Send a task to POST https://api.exepal.com/api/agent/runs with your Exepal key and get the answer as a stream of events or as one JSON reply. It uses the same credit balance as the web app.
Use it from your code
Send your Exepal key. The reply is a stream of events (run, step, tool, tool_result, final); add ?stream=0 to wait for one JSON answer.
curl -N https://api.exepal.com/api/agent/runs \
-H "Authorization: Bearer exk_YOUR_KEY" \
-H "content-type: application/json" \
-d '{"prompt":"Compare the free tiers of 5 uptime monitors","model":"flash","limits":{"max_credits":50}}'
# history, one run, balance
curl https://api.exepal.com/api/agent/runs -H "Authorization: Bearer exk_YOUR_KEY"
curl https://api.exepal.com/api/agent/runs/run_... -H "Authorization: Bearer exk_YOUR_KEY"
curl https://api.exepal.com/api/credits/balance -H "Authorization: Bearer exk_YOUR_KEY"
# follow-up on an earlier run
-d '{"prompt":"Now only the ones under $10/month","continue_from":"run_..."}'
# projects: shared instructions + memory of recent answers
curl https://api.exepal.com/api/agent/projects -H "Authorization: Bearer exk_YOUR_KEY" \
-H "content-type: application/json" -d '{"name":"Competitor watch","instructions":"Prices in USD."}'
-d '{"prompt":"What changed since last week?","project_id":"prj_..."}'
Exepal home · All prices · API docs · Terms · Privacy