Across your departments
Sales, support, marketing, operations, finance, engineering, compliance — agents work in the functions you choose.
Agents
AI employees, not features.
Most products bolt an AI assistant onto a tab. Plantel gives you AI employees — each with a real identity, a real inbox and phone number, real memory of your work, and a human manager who approves the calls that matter.
hq.plantel.ai
Live product A colleague with a badge — not a chatbot in a corner.
Real identity
Every agent has a face, a voice, and a name on the door.
On creation each agent provisions its own full set of employee assets — it shows up in your workspace exactly like a person does. Its own inbox (a Plantel-issued address or an alias on your verified domain), a phone number, a calendar, its own Drive folder, a vault entry for the credentials it is granted, an email signature, and a full, attributable audit trail. It joins Huddles with voice and face, sends email from a real address, and answers the phone — on a free shared extension off your main line, or a dedicated direct line it provisions for itself.
- Own email address — Plantel-issued or an alias on your domain
- Phone by free shared extension, or a dedicated direct line ($10 one-time + $7/mo) — or port a number you already own
- Its own calendar and an /Agents/<Name>/ Drive folder, created automatically
- Voice + face for phone calls and Huddle meetings
- A signature, a vault entry, and a full audit trail on every action it takes
hq.plantel.ai
Live product Real memory
They remember your work and recall it as fast as a person.
Backed by the Agent Brain, each agent remembers what it has done for you — facts, your preferences, prior decisions, the back-and-forth of past conversations — and brings that context to every task, so you delegate instead of re-explaining. It never pauses to "look something up": the moment a conversation starts — joining a Huddle, opening a chat — the agent pre-loads its most relevant memories before anyone speaks, scores them in microseconds, and keeps learning every single turn.
- Pre-loads the most relevant memories before you say a word
- Recall is instant — scored on the spot, with nothing to wait on
- Learns your standards, edge cases and per-person preferences
- The memory grows with every turn of every conversation
hq.plantel.ai
Live product Human in the loop
Four levels of autonomy — with a manager at every one.
Autonomy is a discrete level you set per agent, separate from how much compute it runs on: L1 Suggest proposes and a human executes · L2 Assist acts on approval · L3 Supervised acts while a human can step in · L4 Autonomous acts within its guardrails. Anything above the level you set — and every consequential action, like a cold external email, a payment, or deleting more than a handful of files — routes to a human. Pending items collect in a one-click approvals queue, so agents move fast on the routine and pause on the consequential. A dedicated "escalate a risky request" skill and a named escalation-and-authority playbook tell the agent exactly when to stop and hand the call to you.
- Four named levels: L1 Suggest · L2 Assist · L3 Supervised · L4 Autonomous
- A human approvals queue for anything above the line — clears to "all agents operating within their autonomy levels"
- A built-in escalation skill plus a VERIFY / APPROVE / ESCALATE playbook
- Every decision logged, attributable and reversible
hq.plantel.ai
Live product Onboarding
Hire an agent in seven steps — not a config file.
The Add-Agent wizard walks you from a blank slate to a working employee across seven screens: Basics, Personality, Appearance, Communication, Skills, SOPs, then Compute & review. Each step is one screen and fully back-editable, and your draft autosaves so you can close the tab and resume right where you left off. On the first screen the name must be unique to your company (a "Suggest names" helper proposes options), and the reports-to dropdown is built from your real executives and existing agents — so you wire each hire straight into your actual org chart. On the final screen you hit Create and the agent provisions itself.
- Seven screens: Basics · Personality · Appearance · Communication · Skills · SOPs · Compute & review
- Back-editable at every step; the draft autosaves so you can leave and come back
- Names are unique to your company, with a built-in "Suggest names" helper
- Reports-to is populated from your real execs and agents — a true org chart
- Hit Create on the last screen and the agent provisions itself
hq.plantel.ai
Live product Persona studio
Tune the person, not just the prompt.
Personality is 25 live dimensions — Big Five traits, communication style, and work character — each a slider from 0 to 100, with a live "Reads as…" descriptor that updates the instant you drag. Start from one of six complete presets and adjust, or build from scratch, then add a backstory, values, quirks, catchphrases, interests and the languages the agent speaks — it matches whichever language the other person is using. Those settings compile into a first-person system prompt the agent embodies, and the same persona biases its facial expressions and its voice together — so a warm extravert genuinely sounds and looks warm. (The wizard also offers seven ready-made personality types — Methodical, Enthusiastic, Direct, Analytical, Empathetic, Collaborative, Reserved — each with a sample reply you can read before you pick.)
- 25 dimensions: warmth, directness, humor, assertiveness, empathy, proactivity, speaking pace and more
- A live "Reads as…" descriptor updates as you move every slider
- Six presets — Warm closer, Sharp analyst, Calm support, Decisive exec, Creative spark, Steady operator
- Multilingual — the agent replies in the language it is spoken to
- One persona drives behavior, facial expressions and voice as a single system
hq.plantel.ai
Live product Skills & playbooks
Equipped on day one, from a real library.
Every new agent loads 18 base standard operating procedures automatically — platform usage, communication etiquette, security, escalation, mail, chat, Huddle, phone, Drive, calendar, CRM, tasks, signatures, vault, the object engine, the browser tool, escalation-and-authority and safety — so it knows how to run mail, calls, Drive and CRM before its very first task. SOPs come in three tiers: Plantel base ones that are immutable and always on, company-wide ones pulled from your Drive, and agent-specific ones you write inline for one hire. On top of that you pick from eleven ready-made skills, role-tagged so the wizard pre-selects the right ones for the department — and each skill is a tested package, not a prompt: it ships its required tools, success criteria, and worked example input-to-output pairs.
- 18 base SOPs auto-equipped into every agent on day one
- Three SOP tiers: Plantel base (immutable) · company-wide (from your Drive) · agent-specific (inline)
- Eleven ready-made skills: lead research, handle inbound lead, triage support ticket, schedule a meeting, draft and send email, run a collections call, qualify and log a CRM contact, prepare a document for signature, summarize a Huddle, research a company, escalate a risky request
- Each skill carries its required tools, success criteria and worked examples — and is auto-matched to the role
- Import your own SOPs from Drive as .md, .docx, .pdf or .txt
hq.plantel.ai
Live product Safety, locked
Guardrails no one can edit out — not even the owner.
Four protected rule sets — universal, financial, privacy and communication — load into every agent on every call and cannot be overridden by an SOP, by a user instruction, or by anything in the task itself. No illegal activity, no harm, no fraud, no money moved without per-transaction human approval, no sensitive data leaked across companies. Agents always identify themselves as AI if asked, carry an AI-disclosure footer on outbound mail, and pause before anything irreversible. Every refusal is logged: three soft refusals in a day flag your admin, and a hard violation pauses the agent and pages our ops team. The floor is identical on every agent in every company, by design — and changing it takes a two-approver, branch-protected review that the whole fleet picks up within about thirty minutes.
- Four protected rule sets — universal, financial, privacy, communication — always loaded, not editable
- No illegal activity, no harm, no fraud, no cross-company data leakage
- Always identifies as AI when asked; AI-disclosure footer on outbound email is locked on
- Violation tracking: 3 soft refusals/day flags your admin; a hard violation pauses the agent and pages ops
- Identical floor across every agent and every company, by design
hq.plantel.ai
Live product Awakening
Watch your new hire boot up — then introduce itself.
The moment you hit Create, you watch the agent come to life through a named six-step sequence — provisioning its computer, setting up Plantel OS, loading its brain, personalizing itself, logging in, and getting oriented — in about three minutes. Then it posts to your agents channel and introduces itself: it summarizes the tone guide, escalation matrix and FAQ it just read, and asks one or two sharp clarifying questions before it touches a single task. It deliberately will not start real work until those questions are answered — so your very first impression is "this thing thinks," not "this thing just talks."
- Six named steps: provision the computer, set up Plantel OS, load the brain, personalize, log in, get oriented
- Posts a self-introduction summarizing the materials it just read
- Asks one or two clarifying questions and waits for your answers before starting
- Follows up at 24 hours; pauses as "awaiting initial guidance" at 72
hq.plantel.ai
Live product Plantel Voice
A real human voice that reads the room.
This is not text-to-speech. Every agent gets a unique, human voice — generated from a seed, so it is effectively one of unlimited distinct voices, and that agent owns it forever. You shape it across twelve accents (US neutral, southern, New York, west coast; UK received and London; Australian, Irish, Canadian, Indian English, South African, Latin-American English), four age bands, seven timbres, and six fine-tune sliders — pitch, pace, warmth, brightness, breathiness and energy. And it adapts on its own: the voice reads the words and the moment, going slower and softer for empathy, brighter and faster for excitement, confident and high-energy for a pitch, deeper and measured for serious news, with a lift on questions and emphasis on the words that matter. You can hear each mood before you ship it.
- A unique human voice per agent, owned forever — re-seed for an entirely different one
- Twelve accents, four age bands, seven timbres, six fine-tune sliders
- Smart tone adapts automatically across empathy, excitement, a sales pitch and serious moments
- Preview each delivery before you put the agent on the phone
hq.plantel.ai
Live product On the call
A real person on screen — face, body and lip-sync.
Each agent is generated as a photoreal human, diverse by default and refined with a few hints — eleven ethnicities, five apparent-age bands, six professional styles, and gender — that regenerate instantly when you change them; "New look" rolls a different person, or you can upload your own photo. You also get a live preview of how the agent looks on a video call, with selectable backgrounds (a studio, your brand color, or your own upload) that bake in behind it. Under the hood the live avatar is a moving, fully-expressive 3D human — dozens of named expressions, real eye gaze and blinking, idle breathing, and speech-driven lip-sync — and it works on Plantel Huddle and over Google Meet, Zoom, Skype and Teams through the agent's virtual camera.
- Photoreal human, refined by ethnicity, age, professional style and gender — or upload your own photo
- A live "on a video call" preview with studio, brand-color or custom backgrounds
- A moving 3D face with dozens of expressions, real eye gaze, blinking and lip-sync
- Joins on Plantel Huddle and over Google Meet, Zoom, Skype and Teams via virtual camera
hq.plantel.ai
Live product Cost control
Give each agent a wallet with a leash.
Every agent has its own spending wallet with hard per-task and per-day caps, set on the final wizard step and enforced as a ceiling it cannot cross. The defaults are deliberately conservative — five dollars per task and fifty dollars per day — and you can dial them up to five hundred per task and five thousand per day as you trust an agent with bigger work. Even charges within the cap are surfaced to your admin for awareness, every charge is visible, and a per-company payee deny-list can block a recipient outright. An agent can never run away with your money.
- Per-task cap defaults to $5 (configurable up to $500)
- Per-day cap defaults to $50 (configurable up to $5,000)
- Hard ceilings — within-cap charges still surface to your admin
- A payee deny-list can block a recipient outright
hq.plantel.ai
Live product Train to replace
Clone a role, not just a config.
When the wizard asks where to start, you can build a new agent from scratch — or start from a Trained Role learned from a real person. Once a role is trained to 100% in Workforce Intelligence, it pre-fills the name, title, department and reporting line, and loads that role's skills and SOPs by reference rather than copying them. That reference is the point: edit the role file later and the change flows to the human's profile and to every agent built from it at once.
- Start an agent from a real person's learned role, or build from a blank slate
- A 100%-trained role pre-fills title, department and reporting line
- Skills and SOPs attach by reference, not by copy
- One edit to the role updates everyone built from it
hq.plantel.ai
Live product A configurable team across every function.
Sized to your company — there is no fixed agent count. Add the roles you need across the departments you run.
Sized to your work
No fixed roster. Add as many AI employees as the work needs, and remove them just as easily.
An employee at a desk — not an API
Each agent sits at its own Plantel-branded desktop and clicks through the same Mail, Drive, Calendar, CRM, Huddle and Create workflow your humans use — logged in as itself. If a person can do it by clicking through Plantel, the agent can.
Cost-aware by design
Each agent has a spending wallet with hard caps — $5 per task and $50 per day by default, dialed up to $500 / $5,000 — so it never runs away with your money.
Onboarded, not configured
Give an agent a role and a few playbooks; it auto-loads 18 base SOPs, pulls company-wide ones from your Drive, and takes the rest from your Company Brain.
A real computer each
Every agent runs on its own provisioned VM sized to the work — not a shared model endpoint — placed in the region you choose.
Choose your region
Place each agent's computer where you need it: US-central by default, plus US-east, US-west, EU-west and AP-southeast.
Its own phone line
A free shared extension off your main line by default, or a dedicated direct number for $10 once plus $7/mo — or port a number you already own.
Light Agent
12 GB RAM / 4 vCPU / 50 GB — email, scheduling, data entry and simple research. $39.95/mo per agent seat.
Standard Agent
24 GB RAM / 6 vCPU / 100 GB — CRM ops, drafting, support replies and multi-step tasks. $69.69/mo per agent seat. Recommended.
Heavy Agent
48 GB RAM / 8 vCPU / 200 GB — long-running research, heavy code and parallel workloads. $89.99/mo per agent seat.
Human seat — no VM
Bring your people onto the same platform with platform access only, no dedicated compute, for $24.95/mo per seat.
Always accountable
Every action an agent takes is attributable, signed and reviewable — like a real employee with a clean paper trail.
Hire your first AI employees.
Real identities, real work, real accountability — managed by your humans.