Colrows is a semantic layer platform priced differently from other semantic layer startups: one committed annual figure. Not tokens, not seats, not CPUs, not dashboards. You pay for asking, never for looking. Fix the Context. Not the Model.
Pricing · Updated August 2026
Tokens, credits, seats, CPUs - those are implementation details. They make your invoice volatile and your budget arguments hard to win. Colrows meters one thing: the natural-language requests your people and applications submit.
LLM consumption swings with model choice and prompt shape. Internal model calls, retries, and generated SQL inside a resolution never reach your invoice.
Every tier includes unlimited users. Occasional users and read-only dashboard consumers never count against the meter at all.
No vCPU counting. No environment counting. How you architect your deployment is your decision - and yours alone.
The dashboards, reports, and embedded views those answers feed are included and unmetered. You pay for asking, never for looking.
A question typed into the AI Analyst, or a request resolved through the Semantic Resolution API, is one metered query. That is the meter.
An agent or application call counts once, however many internal reasoning steps it takes. You are charged for asking, once.
Re-running an existing resolution through a dashboard, report, or BI integration is not metered. Only a new question counts.
You sign one number per year. It never changes without your approval.
The line is simple. A natural-language request accepted by Colrows is metered, whichever channel it arrives through. Re-executing an existing answer through a dashboard, report, or BI tile is not.
Each natural-language question submitted to the Colrows AI Analyst chat, including repeated questions.
Each natural-language request accepted by the Colrows Semantic Resolution API.
Each call to the resolution endpoint. One call counts once, however many internal reasoning steps the agent takes.
Dashboard panels refreshing a query that was already authored, and scheduled reports with their recurring refreshes.
Embedded visualisations and BI tiles served from an existing resolution are included, unmetered.
Internal model calls, retries, tool calls, generated SQL, and warehouse sub-queries. Authentication failures, validation errors, and Colrows service failures never consume allowance.
Colrows runs on our infrastructure. We manage the servers, the model, the updates, and the monitoring. You connect your data sources and start asking in plain language. One fixed annual fee.
| Essential | Professional | Business | Enterprise | |
|---|---|---|---|---|
| Best for | Single department | Multiple teams | Cross-department | Full organisation |
| NLP capacity | Sized per tier · annual figure disclosed on request | |||
| Dashboards & reports | Included, unmetered | |||
| Users | Unlimited, no per-seat charge | |||
| Data sources | Multiple | |||
| SSO | Yes | |||
| Uptime target | 99.5% | 99.9% | 99.9% | 99.9% |
| Support | Email + chat | Priority | Priority | 24/7 dedicated |
| On-premise available | No | Yes | Yes | Yes |
| Maintenance | Per governed semantic asset, fixed at contract | |||
| Onboarding | One-time setup engagement | |||
Every tier ships the full compiler: deterministic SQL generation, join-path proof, compile-time governance, persona-scoped access enforcement, audit trails, point-in-time reproducibility, and a usage ledger showing why each request counted. Colrows selects and manages the underlying model on SaaS - no LLM provisioning required.
Each tier includes a monthly capacity of requests. Your allowance is never converted into an overage invoice, and dashboards are never throttled - a single heavy month never disrupts your team.
0 to 80 percent of capacity. Full performance. No action required.
80 to 100 percent. Console banner and email to your administrators. Full performance continues. We reach out to check your capacity is correctly sized.
100 to 110 percent. Full performance continues for the rest of the cycle, so a single heavy month never disrupts your team.
Above 110 percent. New questions pause until the next cycle or an approved query pack. Dashboards, reports, and already-resolved executions continue unaffected.
Add capacity without a full tier move. Packs are priced below your tier's blended rate, co-terminus with your contract, and billed pro rata. Nothing is added automatically - every pack requires your written approval.
Pack sizes are set per tier, and each pack rate is always above the next tier's rate - so a tier move is better value than stacking packs. If your stacked packs approach that point, we will tell you and recommend the move. Enterprise uses a custom volume agreement rather than packs.
For strict data residency, compliance, or air-gap requirements, Colrows deploys directly into your environment. You run the infrastructure and the model. Colrows licenses the Semantic Compiler and supports your team. Data never leaves your environment.
| Tier | Best for | Declared query band |
|---|---|---|
| Team | Smaller regulated teams | Lower annual volume |
| Standard | Departments with steady demand | Mid annual volume |
| Enterprise | Cross-department deployments | Higher annual volume |
| Group | Organisation-wide deployments | Unlimited |
Annual figure per tier disclosed on request. Query bands are set to your customer-declared annual volume and reviewed at renewal - Colrows collects no query telemetry from your environment.
Each tier is the licence plus annual support, combined into one number. No separate deployment licence, no vCPU counting, no per-node or per-environment charge.
Because you run the compute, on-premise is not technically metered. Colrows collects no query telemetry from your environment. The licence tier is selected against a customer-declared annual query band, reviewed at renewal.
Semantic-graph upkeep is included in the annual support bundled into your licence tier. No separate per-asset maintenance charge.
Run your own model - self-hosted, private Azure OpenAI, or your own provider key. Colrows no longer carries your inference cost, because the model runs inside your environment.
AWS VPC, Azure VNet, private GCP, or fully air-gapped. Deploy across as many nodes as your architecture needs. A one-time on-site deployment engagement covers installation, semantic build, security integration, persona mapping, knowledge transfer, and handover. Multi-year schedule fixed at signing, each year capped at 3 percent.
Better value for a given volume. We operate the platform at scale and manage the model for you.
Data residency, air-gap capability, and full control. No metering, no throttling, no telemetry. You choose the model.
Single-year fees are fixed for the year. On a multi-year term the full year-by-year schedule is fixed at signing, each year capped at 3 percent. No year is repriced mid-term.
We pause rather than bill. Your allowance is never converted into an overage invoice, and dashboards are never metered at all.
Every dashboard, report, and embedded view is included and unmetered. Only accepted natural-language requests count; downstream executions do not.
Change model provider at any time without repricing. Your tier is unaffected.
Query packs, tier changes, and capacity-review outcomes are opt-in. Each requires your written approval before it is invoiced.
Pricing and tier definitions are reviewed annually, for new contracts and renewals only. Any change to your renewal is communicated at least 90 days before your term ends.
No hidden line items. Everything below is set out in your order form.
One-time engagement. Covers metadata crawl, semantic graph construction, join-path validation, persona and access mapping, and guided handover. Not repeated at renewal.
One-time on-site engagement by a Colrows forward-deployed engineer: installation, semantic build, security integration, persona mapping, knowledge transfer, and handover. Not repeated at renewal.
Charged per governed semantic asset, billed alongside your SaaS subscription. Keeps metrics, dimensions, entities, and join-path definitions correct as your schema changes. Fixed at contract and renewal. On-premise and VPC deployments receive semantic-graph upkeep as part of the bundled annual support in the licence tier - no separate per-asset charge.
Internal analytics is included in every licence. External-facing applications, commercial redistribution, and customer-facing experiences use an Embedded or OEM agreement, priced per engagement.
Annual in advance, on terms of 12, 24, or 36 months. Invoiced in USD, with AED or INR available on request. An annual capacity review keeps your tier sized to real usage - and nothing changes without your written approval.
Two POC paths, one goal: demonstrate value against your actual data before you commit. The POC also sizes your tier from real query volume, not a guess.
Shared SaaS POC
Access to the Colrows Shared SaaS environment, with guided setup and a walkthrough session. It also sizes your tier from real query volume.
On-Premise / Private VPC POC
Deployment inside your infrastructure or private VPC, delivered remotely by one Colrows engineer. Fee is creditable in full toward your deployment fee on conversion within 90 days of POC close.
Don't find what you need here? Send us a message.
Colrows meters one thing: the natural-language requests your people and applications submit - a question typed into the AI Analyst, or a request resolved through the Semantic Resolution API. Tokens, seats, CPUs, and the dashboards those answers feed are not metered. You pay for asking, never for looking.
Each natural-language question submitted to the AI Analyst chat, including repeated questions; each request accepted by the Semantic Resolution API; and each agent or application call to the resolution endpoint. One agent call counts once, however many internal reasoning steps it takes.
Dashboard panels refreshing a query that was already authored, scheduled reports and their refreshes, embedded visualisations and BI tiles, and everything inside a single resolution - internal model calls, retries, tool calls, generated SQL, and warehouse sub-queries. Authentication failures, validation errors, and Colrows service failures never consume allowance.
No. Every tier includes unlimited users, and there is no cap on how many people use the platform. Occasional users and read-only dashboard consumers do not count against the meter at all, so your real capacity is higher than the headline number.
We pause, we never surprise-bill. Each tier includes a monthly request capacity; we notify you as you approach it, a soft band absorbs a single heavy month at full performance, and above that band new questions pause until the next cycle or until you approve added capacity. Dashboards, reports, and already-resolved executions continue unaffected, and your allowance is never converted into an overage invoice.
Yes. Query packs add capacity at a rate below your tier's blended rate, are co-terminus with your contract, and are billed pro rata in the year they are added. Nothing is added automatically - every pack requires your written approval.
On on-premise and Private VPC deployments, yes. You may run your own model - a self-hosted instance, a private Azure OpenAI deployment, or your own provider key. Colrows no longer carries your inference cost in that configuration, because the model runs inside your environment. On Shared SaaS, the model is managed by Colrows and is not substitutable. Your platform licence and query measurement are unchanged regardless of which model runs underneath, because the Semantic Compiler, join-path proof, and governed SQL generation continue to operate in full.
Yes. For strict data residency, compliance, or air-gap requirements, Colrows deploys directly into your environment - AWS VPC, Azure VNet, private GCP, or fully air-gapped. You run the infrastructure and the model; Colrows licenses the Semantic Compiler and supports your team. Because you run the compute, on-premise is not metered or throttled and Colrows collects no query telemetry - the licence tier is selected against a customer-declared annual query band and reviewed at renewal. Data never leaves your environment.
On a single-year contract your fee is fixed for the year. On a multi-year contract the complete year-by-year schedule is fixed at signing, with each year capped at a hard 3 percent increase. Pricing and tier definitions are reviewed annually, and any change affecting your renewal is communicated at least 90 days before your term ends.
On Shared SaaS, yes - charged per governed semantic asset, billed alongside your subscription. It keeps metrics, dimensions, entities, and join-path definitions correct as your schema and business logic change, and is fixed at contract and renewal so a self-updating graph never changes your fee mid-term. It is separate from platform support: support covers the software, maintenance covers the semantic graph. On on-premise and Private VPC deployments, semantic-graph upkeep is bundled into the annual support included in your licence tier - there is no separate per-asset charge.
Internal analytics usage is included in every licence - your employees, your analysts, your internal dashboards and BI portals. External-facing applications, commercial redistribution, and customer-facing experiences require an Embedded or OEM agreement, priced per engagement based on the application, audience, and distribution model.
Yes - two POC paths. The Shared SaaS POC is free for qualified prospects: one data source connected, the AI Analyst and Semantic API end to end, over 30 to 60 days. The on-premise or private VPC POC is a one-time engagement delivered remotely, with a full semantic build, over 60 to 90 days - creditable in full toward your deployment fee on conversion within 90 days of POC close. On-site deployment is a separate engagement. The POC also sizes your tier from real query volume, not a guess.
Colrows meters the questions you ask - not the model, the servers, the CPUs, the seats, or the dashboards those answers feed. One committed annual figure.
