Lunen.ai is an enterprise agent-building platform that lets a subject-matter expert describe an AI agent in plain language, then hands the resulting plan to IT for approval before it touches any real data. There's no self-serve signup and no public price list — you request early access, get on a call, and go from there.
Short answer: Lunen.ai lets non-technical staff describe an AI agent in plain English while IT sets which tools it can touch and whether it needs human approval before acting. It's early access only — no free trial, no self-serve signup, and no published dollar price. You request access, connect tools like Slack or HubSpot, and pricing gets discussed on a call.
In my testing I went through Lunen's homepage, its early-access request flow, and its pricing page rather than a normal product signup, because that's genuinely all there is right now — Lunen doesn't hand out a free account the way ChatGPT or Claude do. Here's what actually happens when you request access, what the three pricing tiers include, and where this tool is a poor fit versus a strong one.
What you'll need
Lunen is built for teams, not individuals experimenting on a weekend. You'll want at least one of the tools it already connects to — Atlassian, BigQuery, Google Workspace, HubSpot, Slack, or a custom MCP server your company already runs — since an agent with nothing to act on is just a chatbot. You also need someone with the authority to approve an early-access call: this isn't a tool a single employee can quietly adopt, because the whole pitch is that security and IT are in the loop from day one. Come with one real workflow in mind rather than a vague "we want AI." A specific task — "summarize this week's closed deals from HubSpot into Slack" — is what Lunen's early-access team says they want to hear first.
Step-by-step: getting started with Lunen.ai
1. Confirm your tool stack matches
Check whether your company already uses one of Lunen's pre-integrated tools — Atlassian, BigQuery, Google Workspace, HubSpot, or Slack — or runs any MCP server internally. Lunen connects to MCP servers generically, so a tool not on the named list can still work if your team has already wired it for MCP.
2. Request early access
There's no "sign up free" button on lunen.ai — every call to action leads to requesting early access, and the company says the next step is a call to understand your specific AI adoption challenges. This is the biggest departure from a typical AI tool review: you can't just create an account and start clicking.
3. Bring a subject-matter expert and a real task
Once you're through, the core workflow starts with someone who knows the process — not an engineer — describing what they want in plain language. Lunen turns that description into a structured execution plan: named tools, scoped data, and a schedule.
4. Set policy per tool, not per agent
For each MCP tool an agent can touch, someone with admin access decides whether it runs unattended or needs a human to approve every call. This is set at the policy level, so the same rule can apply across every agent using that tool, per Lunen’s own site.
5. Pilot with one team before rolling out further
Given the pricing tiers are built around monthly tool-call volume, a single team's real workflow is a better first test than a company-wide rollout — you'll actually learn what a normal month of usage costs against the included allowance.
6. Review the audit log, then pick a tier
Every action — who approved what, which model ran, what data it touched — lands in a unified audit trail. Once you can see real usage, match it to Operational Control or Enterprise based on your tool-call volume and retention needs.
Example agent instructions you can copy
Because Lunen agents start life as plain-language descriptions rather than typed prompts, what you write here becomes the actual execution plan. A few patterns worth borrowing:
- "Every Monday at 9am, pull last week's closed-won deals from HubSpot, summarize them in three bullet points, and post to #sales — hold for my approval before it sends."
- "When a new ticket is filed in Atlassian tagged 'billing,' check BigQuery for the customer's payment history and attach a one-line summary to the ticket. Run unattended."
- "Once a month, pull the prior month's Google Workspace calendar for the leadership team and flag any meeting with more than 8 attendees for review."
- "If a Slack message in #support mentions a refund, draft a response using our refund policy doc and hold it for a human to send — never send financial commitments unattended."
- "Never pull customer data from BigQuery into a message that leaves our workspace without a named approver signing off first."
Each one names the tool, the trigger, and — critically for a governance-first product like this — whether it runs unattended or waits for a person.
Common mistakes to avoid
The biggest one: treating Lunen like a consumer AI tool you can quietly try solo. There's no free tier and no individual signup, so showing up without buy-in from whoever approves new vendors wastes the early-access call. Second, describing an agent in vague terms ("help with sales stuff") instead of naming the exact tool, trigger, and data it should touch — the plain-language step only works as well as what you feed it. Third, defaulting every agent to unattended mode to save review time, which undercuts the entire reason a governance-heavy team would pick Lunen over a faster, looser tool. Fourth, assuming the 50,000 tool calls a month on Operational Control is unlimited — a handful of chatty agents polling frequently can burn through that faster than expected. Fifth, skipping a real pilot and asking for an org-wide Enterprise rollout on day one, before anyone has audit-log data to justify the annual commitment.
Tools that make this easier
Lunen sits in the agentic AI category, and it's worth comparing against the agent tooling you may already have. If your team is using a general chatbot's built-in agent instead, how to use ChatGPT agents covers that simpler, self-serve alternative. For a different flavor of "AI with guardrails" — a connector that gives Claude standing memory of your other apps — see my Unabyss for Claude setup guide. Lunen isn't the only new agent-building product I've tested this year; Skippr AI takes the opposite approach, embedding a live agent inside your own product rather than orchestrating internal tools. If you're deciding whether any of this is worth it at a smaller scale, best AI tool for small business is a useful gut check before an enterprise governance platform makes sense. I test every tool on this site the same way — hands-on, against the vendor's own site and pricing page — and how we test AI tools explains that process. And if the agents you're building are mainly for engineering workflows, my ranked best AI tool for code guide covers dedicated coding assistants that solve a narrower version of the same problem.
What each plan includes
Lunen doesn't publish dollar prices — every tier past the not-yet-released individual option requires a conversation, per Lunen’s own pricing page.
| Plan | Price | Tool calls | Best for |
|---|---|---|---|
| Solo | Not yet released | Not yet published | Individuals — no release date announced |
| Operational Control | Early access pricing, via call | 50,000/month included | Teams starting their first governed pilot |
| Enterprise | Contact sales, annual commitment | Custom | Org-wide rollouts needing private networking and extended audit retention |
Operational Control includes role-based access control and 90-day audit log retention. Enterprise adds dedicated deployment options, private networking, and longer retention on top of everything in Operational Control, according to Lunen's site.
My take
Lunen.ai is solving a real problem: Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls as the reasons (Gartner, June 25, 2025) — and "inadequate risk controls" is exactly the gap Lunen is built to close. What I'd flag before you request access: there's no published price, no free tier, and no way to try it solo, so this only makes sense if you already have a team and a real workflow lined up. The plain-language-to-execution-plan idea is genuinely clever for getting non-engineers building agents safely. Whether it's worth an annual Enterprise commitment depends entirely on how much your organization already loses to agentic pilots that die in security review — if that number is real for you, Lunen is worth the call. If you're a solo user or small team just exploring what agents can do, there are faster, cheaper ways to start.
Frequently Asked Questions
Is Lunen.ai free?
No. There's no free tier available today — the Solo plan for individuals has no release date or pricing yet. Every other tier requires requesting early access and discussing pricing on a call.
How long does it take to get access to Lunen.ai?
There's no instant signup. You request early access, and Lunen says the next step is a call to understand your AI adoption challenges — how long that takes depends on their team's response time, not anything you control.
What is the easiest way to get started with Lunen.ai?
Confirm you already use one of its connected tools — Atlassian, BigQuery, Google Workspace, HubSpot, or Slack — and bring one specific, real workflow to the early-access call instead of a general interest in AI agents.
Do I need to code to build an agent in Lunen?
No. The core workflow is designed for a subject-matter expert to describe the agent in plain language; Lunen turns that into a structured execution plan with named tools and scoped data.
Is Lunen.ai available to individuals, or only teams?
Right now, only teams. The company says it's "starting with teams on purpose," and the Solo plan for individuals has no release date announced yet.