Plans and Toolkit modules
There are two ways to buy superagnt_. A feature plan gives a team seats and the whole capability set. The builder Toolkit gives a developer a small base and lets the modules arrive one at a time, in session, over MCP. Every price and limit below is read straight from the billing catalog this platform bills from, so what you see here is what checkout charges.
Two purchase shapes
Feature plans are for people running agents. You buy seats and capability together: one price, the full module set, a monthly credit allotment for each wallet, and limits sized to the tier.
The Toolkit is for developers driving the platform from their own client. You start at $19/mo and add the modules you actually reach for. Modules are sold through the MCP server itself: your agent proposes, you confirm on a page, nothing is charged before that.
one fact to carry
Feature plans
The self-serve tiers you can buy in the dashboard. Credits are the monthly allotment included in the price, split across the two wallets; the request rate is per workspace, per minute.
| Plan | Price / mo | Seats | Data credits | AI credits | Rate |
|---|---|---|---|---|---|
| Solo | $99 | 1 seat | $20 | $50 | 120 rpm |
| Team | $199 | 3 seats, then $49/seat | $25 | $100 | 300 rpm |
| Business | $499 | 5 people, then $99/person | $60 | $250 | 600 rpm |
Extra seats add credits as well as capacity, so a bigger team does not immediately run the allotment dry.
| Plan | Per extra seat | Adds data credits | Adds AI credits |
|---|---|---|---|
| Team | $49 / seat | $5 | $20 |
| Business | $99 / person | $10 | $45 |
The counted limits on each tier. A workspace is one superagnt_ and its data; agents are the deployed agents inside it.
| Plan | Workspaces | Agents | Databases | Webhook events / mo |
|---|---|---|---|---|
| Solo | 1 | 5 | 1 | 5,000 |
| Team | 1 | 15 | 3 | 25,000 |
| Business | 1 + 1 per person | 50 | 10 | 100,000 |
Sales-led tiers
These two are provisioned by hand. They render with a contact link rather than a checkout, because there is no price to charge until the terms are agreed.
The Toolkit
The builder base is $19/mo. It carries $10 of data credits, 300 rpm, unlimited connections and custom MCP servers, and every curated data source priced per call. Capability beyond that arrives as modules. If you would rather not think about which ones, All-Access is $139/mo and grants all six at their highest tier, with $25 of data credits and 600 rpm.
| Module | From | Unlocks | Requires |
|---|---|---|---|
| Database & Canvas | $19 / mo | A real Postgres database your agent can query, migrate and load, plus agent-authored canvases and share links.database canvas canvas_share | nothing |
| Automation | $19 / mo | Schedules, queues, and inbound/outbound webhooks.schedules queues webhooks | nothing |
| Data Jobs | $29 / mo | Async bulk pipelines — one cheap structured call per item instead of an agent session each.data-jobs | Database & Canvas |
| Knowledge | $19 / mo | Document collections, ingestion and retrieval.knowledge | nothing |
| Inbox & Messaging | $29 / mo | Email triage, reminders and labels, plus Slack and iMessage bindings. One phone number.email bindings | nothing |
| Agent Runtime | $49 / mo | Deploy, run, invoke and observe hosted agents from your own harness.lifecycle agent-invocation observability | Database & Canvas |
3 of the 6 are tiered (Automation, Data Jobs, Knowledge); the rest are a single price. A rung roughly doubles the price and multiplies the published quantities; the entry rung is what an in-session offer always names.
| Module | Tier | Price / mo | Published allowance |
|---|---|---|---|
| Automation | starter | $19 | 10 schedules · 3 queues · 25,000 webhook events / mo |
| Automation | growth | $39 | 50 schedules · 15 queues · 150,000 webhook events / mo |
| Automation | scale | $79 | 250 schedules · 50 queues · 750,000 webhook events / mo |
| Data Jobs | starter | $29 | 3 active pipelines · 10,000 items / mo |
| Data Jobs | growth | $59 | 15 active pipelines · 75,000 items / mo |
| Data Jobs | scale | $119 | 50 active pipelines · 400,000 items / mo |
| Knowledge | starter | $19 | 3 collections · 500 documents · 2 GB |
| Knowledge | growth | $39 | 15 collections · 5,000 documents · 20 GB |
| Knowledge | scale | $79 | 50 collections · 25,000 documents · 100 GB |
all-access
How gating actually behaves
A module gates its families at call time, not at registration. The tools stay discoverable and the family stays enableable; what changes is what happens when the call lands. Reaching for a family you do not hold returns a requires_upgrade offer as the tool result, carrying the module, its price, whether a trial is still available, and a confirm URL for a human to tap. It is never a protocol error, because clients bury errors and relay results.
{
"status": "requires_upgrade",
"tool": "agnt_schedules_create",
"family": "schedules",
"module": {
"id": "automation",
"name": "Automation",
"tier": "starter",
"price_usd_month": 19,
"requires": [],
"trial_days": 7,
"trial_available": true,
"confirm_url": "https://app.superagnt.com/..."
},
"confirm_url": "https://app.superagnt.com/...",
"human_step": "The call did not run — this tool is sold under the named module and this organization does not hold it. A person has to confirm the purchase: send them the confirm_url; they review the module, its price and trial there. Money is always human-confirmed. A person confirms on the page; nothing here charges anyone. Retry the call once they have confirmed."
}- Any active feature plan — and Toolkit All-Access — grants the full module set at top tier, so those subscribers never see module gating at all. The Toolkit base grants nothing on its own: its modules come from the entitlements you add to it.
- Platform guidance is deliberately never sold under a module. An agent that reads the doctrine succeeds, so the doctrine stays free.
- The workspace database family is exempt from the call-time gate: provisioning is gated instead, and an existing database keeps working through cancellation until the reaper says otherwise.
- A module that requires another (Data Jobs and Agent Runtime both require Database & Canvas) bundles the dependency into the same confirm action.
Trials
A trial is card up front and once per organization, ever. It is a real Stripe subscription with a trial period, which is why deploy entitlement and plan limits behave normally during it, and why nothing breaks at conversion. Trialable tiers are solo, team, business, toolkit.
read the length, do not assume it
trialDurationDays from GET /dashboard/billing/entitlement rather than printing a constant, or your badge will quote a trial the checkout does not grant.What a feature-plan trial seeds, per tier. The data credits arrive as a one-time grant on the purchased balance, so they survive the trial. The AI credits are the trialing subscription's allotment, so they expire with it and swap to the full allotment on conversion.
| Plan | Trial data credits | Trial AI credits |
|---|---|---|
| Solo | $10 | $50 |
| Team | $25 | $75 |
| Business | $30 | $100 |
The builder trial steps down instead of switching off
- 01
Days 1 to 7: fully equipped
A Toolkit trial holds every module at its top tier. You are not choosing modules while you are still working out what you need.
- 02
Day 7: auto-convert to the base
The subscription converts to the $19/mo base. The full kit stays live through a 3-day grace window, so the first paid days are not half a product.
- 03
The cut: offers become itemized
When the grace window closes, a sweep records which modules saw real use and consumes those modules' once-ever trials. From then on, a gated call returns an offer that lists exactly what you used, what each costs to restore, and the All-Access price to keep everything. All-Access is deliberately not a trial target; it is what the cut upsells to.
Module trials are separate and smaller: 7 days, once per organization per module, and they need an active Toolkit subscription because a trial mints against a card already on file. An organization with no plan gets offered the bundled checkout instead.
Upgrading from inside an agent session
Three control tools cover the whole path from "what are we paying for?" to a page a person taps. None of them moves money, on any surface.
| Tool | What it does |
|---|---|
agnt_subscription_get | Read-only billing state: the current plan, every Toolkit module as owned, trialing or available with its tier and price, the credit commitment, and seats. |
agnt_subscription_preview | Dry-runs a plan change against Stripe: what switches and what proration would be invoiced right now. Returns an idempotency token, which is the only way to proceed. |
agnt_subscription_change | Turns that token into a confirm URL for your user. The change executed is exactly the one previewed. Money moves only when a person confirms on the page. |
Valid targets for a preview or change are solo, team, business, toolkit, toolkit_all_access. Grandfathered and sales-led plans can be the source of a change but never the destination.
What is enforced and what is copy
Five resources are counted and enforced. Crossing one raises PLAN_LIMIT_REACHED with HTTP 402 and a message naming the resource, the usage and the cap: workspaces, agents, databases, webhook events and seats. Runtime hours exist as a countable resource but no current plan caps them.
Toolkit module tier quantities are different. The schedule, queue, pipeline, collection and storage numbers in the tier table above ship as published allowances, not hard caps. Holding the module is the gate; the quantities are what the tier promises, and enforcement is deliberately deferred so that publishing them is a grace period rather than a price increase.
toolkit base caps come from modules
Rate limiting is separate from plan limits: the rpm figure is a per-workspace budget per minute, and exceeding it returns 429 with a Retry-After header. See Errors and rate limits for the full error surface, and Usage for where spend is recorded.