What the agent thinks through.
Every chat session, voice minute, email reasoned, SMS conversation. The cognitive cycle is the unit. Predictable, transparent, billed by the channel’s natural rhythm.
KARMAFLOW PRICING · MAY 2026
Karmaflow pricing is built around deployed agents, channel-based reasoning, visible actions, and custom-scoped intelligence work — so you can understand what you are paying for before your agents go live.
Custom-scoped intelligence work starts at $2,000+ — for operations agents, intelligence agents, compounding brain, and deep work engagements.
No platform fee tier games. No seat counts. No surprise overages.
THE PRICING MODEL
Karmaflow’s rate card has three clearly-separated layers, in the order you’d encounter them in a real deployment. Most customers live in the first two; the third is reserved for work that genuinely doesn’t fit a metered unit.
Every chat session, voice minute, email reasoned, SMS conversation. The cognitive cycle is the unit. Predictable, transparent, billed by the channel’s natural rhythm.
CRM updates, sends, transfers, memory operations, image work, research calls. Actions are layered on top of reasoning — honest, visible, and only charged when actually performed.
Operations agents, Intelligence agents, Deep Work Engagements. Cognitive shape is unpredictable, so the work is scoped — not metered. A claims agent runs 4 steps for one case and 200 for another. We honour that variance with custom contracts.
scope, not meter
See custom-scoped work →Predictable where it should be. Custom where it has to be.
LAYER 1 · REASONING FEES
These rates apply when the channel is the agent’s primary work — a chat agent answering inquiries, a voice agent handling calls, an email agent running prospecting. Same agent reasoning principle, billed by each channel’s natural unit.
per sessionresets at 30 min idle
per email handledread or written
per conversationresets at 30 min idle
per sessionresets at 30 min idle
per minuterounded up
Whether the customer says one sentence or fifty,
whether the conversation lasts two minutes or two hours —
the reasoning happened once. So the price is once.
SAME LOGIC, DIFFERENT UNIT — VOICE IS REAL-TIME, METERED BY TIME, NOT EXCHANGE.
LAYER 2 · ACTION RATES
Reasoning is the work. Actions are the consequences — sending an email, updating a CRM record, transferring to a human, generating an image, executing a workflow tool.
Actions are charged separately because they’re discrete, observable, and predictable. A session that involves no actions costs only the reasoning fee. Actions are extras, but they’re not surprises — they’re a named layer of the bill, with rates the customer chose by configuring their workflow.
Honest extras. Not hidden fees.
Create or update a record in your connected CRM.
Email sent by an agent during another agent's session.
SMS sent by an agent during another agent's session.
Agent reads from or writes to working memory or knowledge graph.
Agent analyzes a customer-provided or system image.
Agent generates an image as part of its reasoning.
Execute a custom MCP tool, trigger a workflow, or call a connected system.
Agent briefs the human teammate before handoff. Voice minutes during the handoff billed at $0.35/min.
Agent routes the conversation to a human and disconnects.
When email is the agent’s primary work (an Email Agent doing prospecting, autoresponse, or nurture), the agent does full read+write reasoning — interprets, evaluates, drafts, threads memory. That’s a $1.00 reasoning cycle.
When email is sent by another agent during its session (your voice agent emailing a confirmation), the parent agent has already done the cognitive work. The email is a bounded output action, not a separate reasoning cycle. Lower token cost, lower price.
Same logic applies to SMS-as-action vs. SMS-as-channel.
CROSS-SESSION MEMORY · OPT-IN
Most agents forget the customer the moment the conversation ends.
Karmaflow agents remember.
When Cross-Session Memory is enabled, every agent recognizes returning customers and threads context across sessions. Every interaction draws on the full history of the relationship — across every channel, across every prior session, across the full lifetime of engagement.
This is concierge-grade context, automatically, for every customer. Personalization that was never possible at scale before AI agents.
$0.25 per engagement · opt-in
Per chat session, voice conversation, SMS session, WhatsApp conversation, or email — only when memory is invoked.
LAYER 3 · CUSTOM-SCOPED
Some agents do work that can’t be priced per session, per minute, or per action — because the cognitive shape is genuinely unpredictable. We honour that with scoped contracts, not metered surcharges. Three categories, all custom by architecture, not by accident.
$150/month + custom usage by contract
Multi-step back-office workflows on a single domain — claims processing, onboarding, AR follow-up, mediation, document validation, fraud review. A claims-processing agent might run 4 steps for a routine claim, 40 for a complex one, 200 for a contested case with multiple parties. The variance is the distribution, not the tail. Per-action pricing would distort the relationship.
$150/month + custom usage by contract
Scheduled or on-demand analysis using the knowledge graph and graph data science — trend detection, churn modeling, segmentation, supplier analysis, fraud pattern detection. Cognitive load varies with cadence, query complexity, and data volume. These agents consume heavy compute, produce strategic outputs, and steer business decisions.
Custom-scoped per engagement
Multi-agent workloads producing strategic outputs across systems — audits, document synthesis, cross-channel research, compliance reviews, multi-system reconciliation. Operations and Intelligence agents are standing capabilities — they run continuously. Deep Work is engagement-shaped — it produces a specific deliverable. Multiple agents collaborate to produce one unified output.
If the work has unpredictable cognitive shape,
it doesn’t get a unit price. It gets scoped.
COMPOUNDING BRAIN · TENANT-WIDE
Knowledge graph. Graph data science. Persistent learning.
Every interaction, every analysis, every correction — sharpens the next.
Compounding Brain is the tenant-wide intelligence layer that turns Karmaflow from a workforce of agents into a learning organization. Every conversation can write back to the graph. Every Data Science agent reads from one consolidated source of truth. Every audit reasons against the full history.
The Brain isn’t required for engagement work. But for any deployment where intelligence compounds — pattern detection, churn modeling, supplier analysis, multi-channel synthesis — it’s the foundation.
Starting at $2,000/month
Tenant-wide. Always custom-scoped to your deployment.
PRICING CALCULATOR
Pick the agents you want, set realistic monthly volumes, watch the price update in real time. No sign-up needed. No data captured.
EVERYTHING ELSE
Beyond the rate card, here’s how integrations, onboarding, hardware, residency, and support work. Most are included or starting fees — no hidden enterprise tiers.
For data-sensitive verticals — healthcare, legal, government, regulated finance — that need on-premises or hybrid deployments. From $7,000 + hardware, scoped by complexity.
TRUST · GOVERNANCE · PRIVACY
Every Karmaflow agent includes built-in review and compliance — at no extra cost. Every conversation is audited, every action is verified, every policy boundary is enforced. This isn’t an add-on. It’s how Karmaflow agents work.
See our safety record →Cross-Session Memory is keyed to the customer — not stored as PII. Identifiable details live in your source systems. Karmaflow’s memory layer references them by key. Architecture, not policy.
Read our security architecture →Every agent decision, every action, every escalation flows to the Unified Action Ledger — with actor, channel, timestamp, and provenance. Reconstructible by design.
See governance details →FAQ
Yes. You pay $50/month for the platform, $150/month per agent you deploy, and usage rates only when usage happens. No seat games, no tier games, no minimum commits. You can scale up or down month to month.
A chat session is the conversational thread between the customer and the agent. It resets after 30 minutes of inactivity. There is no upper bound on turn count — a 50-turn conversation within continuous engagement is one session at $1.00. The same rule applies to SMS and WhatsApp.
When email is the agent's primary channel — your Email Agent doing prospecting, autoresponse, or nurture — it's $1.00 per email handled. The agent does full read-and-write reasoning for every message. When email is sent by another agent during its session — your voice agent emailing a confirmation — it's $0.25 per email. The parent agent has already done the reasoning; the email is a bounded output action.
Every Karmaflow agent runs with built-in review and compliance. Every conversation is audited. Every action is verified against your configured policies. Every escalation routes to the right human with full context. It's part of how agents work — included in the $150/month agent fee, not sold as a separate product.
Because the work has unpredictable cognitive shape. A claims-processing agent might run 4 steps for a routine claim, 40 for a complex one, and 200 for a contested case. A Data Science agent's compute load varies with cadence and complexity. Trying to meter this work per-action would distort the relationship — sometimes punitive, sometimes underpriced. We honour the variance with scoped contracts.
Cross-Session Memory makes agents recognize returning customers and thread context across sessions. When enabled, a $0.25 charge applies to each engagement (chat session, voice conversation, SMS session, WhatsApp conversation, or email) where memory is invoked. Opt-in — only charged when used. Memory is keyed to the customer, not stored as PII.
Voice is billed per minute, rounded up. A 31-second call is 1 minute ($0.35). A 5-minute-and-1-second call is 6 minutes ($2.10). Industry-standard, customer-expected.
No. Volume discounts are available for high-usage deployments — but they are discussed transparently with our team. Hardware deployments have a "+ hardware" line item that is customer-owned. Operations, Intelligence, and Deep Work are scoped per deployment — but always with predictable monthly investment, never surprise overages.
Three paths. No pressure.
Build your workforce, see your monthly cost in real time. Save the configuration, share with your team, come back when you’re ready.
Open the calculator →Walk through your use case with us. We’ll size the agents, recommend integrations, and build a tailored quote. 30 minutes, free, no obligation.
Book a call →Spin up a demo tenant, build a real agent, run real interactions. We’ll help you set it up — included.
Request demo access →Karmaflow is built for non-coders, C-suite, and managers.
Most businesses are running their first agent within a week.