- BFSIBanking, Financial Services & Insurance
- HealthcareComing soon
- ManufacturingComing soon
- Retail & E-CommerceComing soon
Snowflake CoCo
Your data engineers paste table names into a chat window. The agent should already know your schemas — Snowflake CoCo, formerly Cortex Code, drafts SQL, dbt models and Airflow DAGs from your account’s schemas, roles and policies — in a desktop app, a CLI, Snowsight or your IDE, billed on tokens rather than seats.
Buy through TechBag
Same software. Better outcome — at a lower cost.
How it’s rated
Full scoreboard ↓Quick answer
The Snowflake platform family
This page covers Snowflake CoCo — the AI coding agent for data work, formerly Cortex Code. The rest:
Quick facts
30-second orientation- Product
- An AI coding agent for data engineering that works with your Snowflake schemas, roles and policies as context
- Maker
- Snowflake Inc. (NYSE: SNOW), led by CEO Sridhar Ramaswamy since February 2024
- Name
- Formerly Cortex Code; Snowflake’s old Cortex Code product page now redirects to Snowflake CoCo
- Surfaces
- A desktop app, a CLI and Snowsight, the Snowflake web interface, plus extensions for IDEs
- Price
- Model tokens billed in AI credits: $2.00 per AI credit at the Global rate, $2.20 Regional
- Standalone
- A CoCo CLI subscription that needs no Snowflake workloads, as CRN Asia reported; price not published
- Tooling
- dbt and Apache Airflow support, announced in Snowflake’s own video in February 2026
- Adoption
- More than 9,100 accounts in the quarter to 31 July 2026, by Snowflake’s results release
- India
- The account can live in AWS Mumbai or Azure Central India (Pune); model-call location not documented
- In India via
- TechBag — token budget model, quote in INR with GST, a pilot on one real pipeline
Understand data-aware coding agents before you buy one
Most product pages skip this. We start here — so you buy a capability, not a buzzword.
What is a data-aware coding agent?
An AI agent that writes SQL and pipeline code with your warehouse’s schemas and permissions already in view.
A general assistant with pasted schemas vs a data-aware agent — the honest table
What consolidation actually replaces, dimension by dimension.
| Dimension | General assistant, schemas pasted | Snowflake CoCo |
|---|---|---|
| Where table names come from | Pasted into a chat, or guessed | Read from the account’s schemas |
| Permissions in the draft | Found out when the query fails | Roles taken in as context |
| How it is paid for | A seat for every engineer, used or not | Tokens in AI credits, per model |
| Pipeline code | SQL only; dbt and DAGs by hand | SQL, dbt models and Airflow DAGs |
| Where engineers use it | One editor plug-in | Desktop, CLI, Snowsight or an IDE |
| What it is NOT | — | A general coding assistant or a fixed-price seat |
The cheapest test is the trial credits: point CoCo at one dbt project, log the tokens per task, and price a month from that.
The five pieces of the platform
Vendors love diagrams; buyers need to know what they’re actually operating. Here’s the whole platform, demystified.
Context
CoCo draws on the schemas in your account, its role-based access and its governance settings, so a draft query names tables and columns that exist rather than plausible guesses.
Surfaces
The same agent sits in a desktop app, a command-line tool, the Snowsight browser interface and IDE extensions, so a data engineer reaches it from whichever window is already open.
Models
Every request is counted in tokens and charged in AI credits per million at the rate Snowflake lists for that model; the consumption table carries those rates and changes month to month.
Plans
Inside a Snowflake account it draws on AI credits. CRN Asia reported a separate CoCo CLI subscription for teams that do not run workloads on Snowflake; its price is not published.
An agent that reads your Snowflake context — schemas, roles and policies — then drafts code wherever you work, billed per token.
Nine capabilities. Context, build, govern.
Snowflake CoCo writes data code with your Snowflake account’s schemas, roles and policies already in view.
Starts from your real tables
Drafts begin from the databases, schemas and columns in your account, so fewer queries fail on a misspelt or invented name.
Aware of who is asking
Snowflake lists role-based access among the context CoCo uses, so the grants on your role shape what it proposes to touch.
Policies are part of the brief
Governance settings sit beside schemas and roles in CoCo’s context; check in a pilot how your masking rules appear in its drafts.
Desktop, terminal or browser
Use it in the desktop app, from the CLI next to git, inside Snowsight, or through an IDE extension; the agent is the same.
Works inside dbt projects
dbt support, announced in February 2026, lets the agent help with models and tests in a dbt project as well as plain SQL.
Orchestration code in scope
Apache Airflow support came in the same update, so DAGs that schedule the loads can be drafted beside the transformations.
No seat licence to count
Inside an account CoCo is billed on token consumption, charged in AI credits at a separate rate for each model it uses.
Free credits for new users
Snowflake gives new users trial access with free credits, so a team can measure token use on its own code before it budgets.
Strong authentication by default
Snowflake completes its move off password-only sign-ins in October 2026; connect the CLI with key-pair, OAuth or a token.
Watch Snowflake CoCo in action
A customer story from Inogen, the data development lifecycle in CoCo, its June 2026 launch, and the dbt and Airflow update made under the Cortex Code name.
How Inogen Cuts Development TimeIn Half With CoCo
A Snowflake customer, Inogen, on using CoCo in its data development work; the halved time is the customer’s own claim.
Simplify The Entire Data Development Lifecycle with Snowflake CoCo
Snowflake’s walk through CoCo across the data development lifecycle, released with the new name.
Snowflake CoCo: Welcome to the Agentic Enterprise
The launch framing for CoCo as Snowflake moved its agents to new names in June 2026.
Cortex Code just changed the data game. Now with dbt and Airflow support! #cortex
Recorded under its former name, Cortex Code: the update that brought dbt and Apache Airflow support.
Want a live, India-context walkthrough for your environment?
Book a guided demo →Coding assistants see the file, not the warehouse. CoCo starts from your Snowflake account.
Here’s what genuinely sets it apart — and exactly where it stops.
It starts from your account, not a blank file
A general coding assistant sees the code in your editor; table names, grants and policies have to be pasted in or guessed. CoCo is built to take your Snowflake schemas, role-based access and governance settings as context, so a first draft of a query or a dbt model refers to objects that exist. That is the gap it closes for data engineers.
The bill follows use, not headcount
There is no seat fee inside an account: CoCo is billed on tokens, in AI credits at $2.00 Global or $2.20 Regional. Model rates differ; claude-sonnet-4-6 is listed at 1.65 AI credits per million input tokens and 8.25 per million output, about $3.30 and $16.50 at the Global rate (our arithmetic). Light users cost little; heavy ones do not.
Where data engineers already work
The agent runs in a desktop app, a CLI, Snowsight and IDE extensions, and since February 2026 it handles dbt and Apache Airflow code too. CRN Asia reported a standalone CLI subscription, Snowflake’s first, for teams that do not run workloads on Snowflake. Snowflake counted more than 9,100 accounts using CoCo by 31 July 2026.
Where it stops
It is built for data work, not general application code, so most teams still keep a coding assistant for the rest. The standalone CLI price is not published. Token billing has no ceiling of its own, and switching models can move the bill a long way. Where its model calls are processed is not documented for India, and no analyst has rated it.
The numbers behind the platform
What your Snowflake CoCo rollout looks like
Pick the pipeline to test on
Choose one real dbt project or DAG with known pain, and name the engineers and roles who will use the agent on it.
Fix sign-in before the pilot
Move any password-only service users to key-pair, OAuth or a token, since strong authentication is enforced in October 2026.
Run it on trial credits
Use the desktop app, CLI or Snowsight on the chosen work and log tokens per task by model, from the account’s usage data.
Turn tokens into a budget
Multiply measured tokens by each model’s AI-credit rate, set alerts on AI credits, and agree which models are allowed.
Widen to the data team
Roll out to the rest of the data engineers, review drafts as you would a colleague’s, and recheck the model rates monthly.
The review scoreboard
Modelled on Gartner Peer Insights structure. *Counts and breakdowns are illustrative pending verified review collection.
Quick poll — what’s driving your evaluation?
Talk to an advisor“It wrote a join across three schemas using the right column names first time. Our old assistant kept inventing them.”
“We rebuilt a dozen dbt models with it in a sprint. Review still matters, but the boilerplate tests came out right.”
“Running it from the terminal next to git suits our team better than the browser. The Airflow DAG drafts saved real time.”
“Token billing was cheap for most of us; two heavy users drove the bulk. Put a budget alert on AI credits early.”
“Our service user still had a password. The CLI connection failed until we moved it to key-pair authentication.”
“Great on Snowflake SQL, less useful for the Java services around it. We still pay for a general coding assistant.”
Where everyone sits — the grids
Analyst firms bury this view behind paywalls, and G2 retired its Grid. So here’s TechBag’s synthesis of the AI coding assistant market — tap any vendor to see why it sits where it does.
TechBag AI Coding Assistant Grid
Execution strength vs product vision — the classic market map, minus the paywall.
9,100+ accounts; tokens billed in AI credits.
Data Context × Coding Breadth
The grid nobody publishes — how much of your data platform the agent understands vs how many surfaces and kinds of code it covers.
Schemas, roles and policies; desktop, CLI, Snowsight, IDE.
Positions are TechBag’s illustrative synthesis of public review-platform data and vendor documentation — not a reproduction of any analyst graphic. Verify before relying on it.
Snowflake CoCo vs the AI coding assistants
Against GitHub Copilot, GitLab Duo, Cursor, Databricks Genie Code and Amazon Q Developer — on data context, surfaces, price, allowances, governance and India.
| Dimension | Snowflake CoCo | GitHub Copilot | GitLab Duo | Cursor | Databricks Genie Code | Amazon Q Developer |
|---|---|---|---|---|---|---|
| What it is | Data-work coding agent | General coding assistant | AI inside GitLab | AI-first code editor | Workspace data agent | AWS coding assistant |
| Where it runs | Desktop, CLI, Snowsight | IDE, CLI, app, web | GitLab UI and IDEs | Its own editor | Databricks UI only | IDEs, CLI, AWS console |
| Data-platform context | Schemas, roles, policies | Your codebase, not data | Lifecycle, not data | Repo context only | Unity Catalog grants | Natural-language queries |
| Pricing model | Tokens in AI credits | Per seat plus AI credits | Tier plus GitLab Credits | Per user, monthly | DBUs past an allowance | Per user, monthly |
| Published entry price | $2.00 per AI credit | $19 per seat a month | $29 Premium seat | $40 per user (Teams) | 150 free DBUs a user | $19 per user (Pro) |
| Included vs extra | Every token is billed | 1,900 credits a seat | $12 credits, promotional | Limits not quantified | Compute billed apart | Pro limits unstated |
| Governance and access | Inherits account RBAC | SAML SSO, Enterprise | Inside GitLab’s controls | SSO, SCIM, audit logs | Unity Catalog grants | IP indemnity on Pro |
| Pipelines and tools | dbt and Airflow | Any repo, any language | GitLab CI pipelines | MCPs, skills, hooks | Lakeflow, notebooks | AWS data services |
| Self-hosting | SaaS only | SaaS only | Self-managed GitLab | Cloud service | Inside Databricks | AWS-hosted |
| India storage region | Account in Mumbai/Pune | Not documented | Your self-managed site | Not published | Geo-based residency | Not verified |
| Free tier and trial | Trial credits | Free: 2,000 completions | Ultimate trial | Hobby plan, free | 150 DBUs every month | Free: 50 agent requests |
| Adoption signal | 9,100+ accounts | Most-adopted assistant | GitLab installed base | Not published here | Not published here | Not published here |
| Lock-in and exit | Tied to Snowflake data | Portable code | Tied to GitLab | Editor switch | Tied to Databricks | Leans on AWS |
| Best fit | Snowflake data teams | Whole-company coding | GitLab estates | AI-first developers | Databricks data teams | AWS-centred teams |
Which approach fits you?
Honest fit signals — because the fastest way to lose your trust is to pretend one product wins every scenario.
Choose Snowflake CoCo if…
- ✓Your data engineers write SQL, dbt models and Airflow DAGs against Snowflake every day
- ✓You want drafts built from your real schemas, roles and policies rather than pasted context
- ✓You would rather pay per token than buy a seat for every analyst who might try it
Compare alternatives if…
- ✓Most of your code is application code — GitHub Copilot, Cursor and Amazon Q Developer are general assistants
- ✓Your code and pipelines live in GitLab — Duo comes with Premium and Ultimate credits
- ✓Your data platform is Databricks — Genie Code works from Unity Catalog the same way
Do not expect…
- ✓A published price for the standalone CoCo CLI subscription
- ✓A self-hosted edition, or documented Indian processing for its model calls
- ✓A spending cap by default — set budgets on AI credits yourself
Snowflake CoCo is one of 35 developer tools products TechBag carries. The Developer Tools guide narrows them to a shortlist and shows the reasoning. →
What does hand-written data plumbing cost you?
Drag the sliders (pipelines and dbt models maintained; data-engineer hour cost). Estimates model engineering time spent looking up table names, fixing permission errors and writing boilerplate SQL, tests and DAG code at an assumed 1.5 hours per pipeline or model a year, with 70% of it removed by an agent that drafts from your account context. Both figures are assumptions, and token costs are not included. Illustrative.
Loaded cost = salary + overheads per productive hour. Illustrative only — your TechBag quote models your actual environment and modules.
Three ways to consume it
Published: inside a Snowflake account CoCo has no seat fee — it is billed on model tokens in AI credits, $2.00 each at the Global rate or $2.20 Regional, at per-model rates in Snowflake’s consumption table (claude-sonnet-4-6: 1.65 AI credits per million input tokens, 8.25 per million output). New users get trial credits. A standalone CoCo CLI subscription has been reported, but its price is not published. TechBag measures your tokens in a pilot, then quotes in INR with GST.
In-account CoCo
Best for teams already on Snowflake
- Tokens billed in AI credits, no seat fee
- $2.00 per AI credit Global, $2.20 Regional
- Trial credits for new users
+ Platform add-ons
Best for a broader rollout
- Scoped to your estate
- Add-on modules as needed
- Phased, right-sized deployment
Standalone CoCo CLI
Best for teams not running Snowflake workloads
- Subscription reported by CRN Asia
- Price not published; ask in writing
- dbt and Apache Airflow support
Buy it for less — TechBag pricing beats list
Whatever the list prices above, TechBag negotiates a significantly better deal — with GST-compliant INR invoicing and local support. Ask us for your discounted quote.
Get an India-ready quote
Tell us your requirements and current tools — we’ll model it against what you spend today.
The 8 questions to ask every vendor
Take this into your next vendor call — including ours.
Is most of the work SQL, dbt and Airflow on Snowflake? Application code is better served by a general assistant.
In-account token billing or the standalone CLI plan? Ask Snowflake for the standalone price in writing.
Which models will you allow? Rates per million tokens differ widely in the consumption table, and change monthly.
Who watches AI-credit spend, and what alert fires before a heavy user runs through a month’s budget?
Are any service users still on passwords? They stop working once October 2026 enforcement completes.
Do the roles granted to engineers match what the agent should see? It works from those roles as context.
Is the account in AWS Mumbai or Azure Pune, and has Snowflake stated in writing where CoCo’s model calls run?
Who reviews agent-written SQL and DAGs before they reach production, and is that step in your pull-request rules?
Questions buyers ask
Ready to evaluate Snowflake CoCo?
Measure tokens on one real pipeline first, or let a TechBag advisor scope a pilot that turns that usage into an AI-credit budget by model.
Stats, ratings, review counts and pricing are illustrative and sourced from public materials; verify before purchase.