DevelopersBusinessPricingBlog

Most clouds hand you a runtime. We cover the whole life of the system.

Data, CI, registry, compute, sign-in and models on one platform instead of five vendors: one bill, no seats, running in the UK/EU. Drive all of it from the CLI or your agent over MCP.

Gitrepos, branches, PRs
Git CIworkflows on every push
Registryimages and artifacts
Igniteapps, APIs, functions
AppIDusers, roles, SSO
AI Modelsinference and embeddings
DataK³SQL, vectors, graph, objects
one org · one set of permissions · one billbuilt for agents · every station, over MCP
The product range

The whole range, on one copy of the data.

The parts of a running shop. Pick one to see what it does.

DataK³
One bucket: a row is a table row, a vector and a graph node, and its files sit beside it.
Postgres wirepgvectorCypherS3 objectsACID
Open DataK³
one bucket · shop
42 · Yoga mat · 34.00 · [0.12, …] · 42.jpg
SQL · a row
41Mat strap
42Yoga mat
43Cork block
vector · <=>
graph · bought_with
objects · S3
products/
42.jpg
data_object_url → signed link
one statement: SELECT sum(qty * price) … WHERE product_id IN (SELECT node FROM graph_khop('shop', 42, 2))
How it connects

How the parts talk to each other.

The same shop, running. A push becomes a rollout, a request crosses the gateway into one ACID commit, a search ranks vectors and walks the graph, and your agent operates all of it. Every edge is a protocol you already speak.

>_Your agent
Git forge
{ }Clients
Registry
AppID gateway
Ignite
Models
DataK³
orders-apisearch-apirecs-worker
SQLvectorgraphobjects
git push
build · test · push image
deploy.yaml → orders-api v8
01 · push → live

Merge to main, and it is live: CI on the forge, an image in the registry, a rollout on Ignite.

pick any part on the map to see its interface
SpeaksPostgres wire · pgvector · Cypher / Bolt · S3 · OCI · GitHub Actions format · OpenAI API · MCPRead the docs
How you work

From an empty account to the running shop.

Connect your agent once, then build what the map shows with Claude in your editor: model the bucket, write the services, push and ship, operate. Every call is a real dodil tool.

ModelYou ask for the data model. Claude writes the schema and creates it in one bucket: Postgres tables, a vector column and a graph over the same rows.
schema.sqlshop
schema.sql
services/orders/main.py
services/search/main.py
services/recs/worker.py
.github/workflows/ci.yml
.dodil/deploy.yaml
ops/notes.md
1CREATE TABLE products (
2 id BIGINT PRIMARY KEY,
3 title VARCHAR,
4 price DECIMAL(12,2),
5 embedding VECTOR(768)
6);
7CREATE TABLE order_lines (order_id BIGINT, product_id BIGINT,
8 qty INT, price DECIMAL(12,2), PRIMARY KEY (order_id, product_id));
9CREATE TABLE bought_with (src BIGINT, dst BIGINT, PRIMARY KEY (src, dst));
10
11CREATE INDEX ON products USING hnsw (embedding vector_cosine_ops);
12CREATE GRAPH shop NODES (products KEY id) EDGES (bought_with SRC src DST dst);
Claude dodil MCP
model a shop: products, orders, what sells together
data_bucket_create shop
data_table_query CREATE TABLE products …keys and types enforced
data_table_query CREATE INDEX … hnswdistances re-ranked exact
data_table_query CREATE GRAPH shop …
✓ one copy: OLTP, warehouse, vectors, graph
↺ the same verbs from the CLI, the console or any MCP agentConnect your agent
Why Dodil

Fewer vendors, a smaller bill, and your data where it has to stay.

What the shop above costs you to run, and what it never ties you to.

1bill, not five vendors

Data, CI, registry, compute, sign-in and models under one org, one set of permissions and one invoice, with nothing to glue between them.

the map above is the whole stack
0seats

Metered by storage, compute and tokens, with a free allowance on every meter and hard caps. Agents and CI runners never count as people.

Free $0 · Pro $19 · Scale $99, same unit rates
UK/EUwhere your data stays

On hardware we own, with London live today. The models run there too, next to the rows they read.

the globe below
Opendoors, no lock-in

Postgres wire, S3, OCI, GitHub Actions format, the OpenAI API and MCP. Your code and data export any time, and egress is free.

bring your own client, driver or SDK

From the silicon up —
hosted in London, for the UK/EU.

We own the entire stack — our own silicon, servers and network, with knowledge, compute and inference layered on top — all in one London region. One vendor from the metal to the model; your data stays in the UK/EU.

Owned end to end
Servicesyour apps & agents
InferenceAI Models
ComputeIgnite
KnowledgeK3
Serversour bare metal
Siliconour GPUs
London regionUK · data stays here
London · liveFrankfurt · Middle East · Africa — planned
SOC 2in progressISO 27001in progressGDPR-readyData residency enforced
Pricing

Pay for what runs.

One rate card, three simple plans — start Free with a monthly allowance on every meter and no card required (usage stops at the caps, never a surprise bill). Pro ($19/mo) and Scale ($99/mo) pay the same per-meter rates below; the fee buys capabilities and lets usage keep going past the allowance.

See the plans & unit rates →
Scales to zero

Pay for bursts, not for idle.

Ignite Apps and K3 Tables compute autoscale pods down to zero between requests. AI Models, vector reads, storage — everything else is metered per token, per GB, per request.

VM baseline
You payIdle — free
IgniteAI Models
Per tokenhosted LLMs · × per-model multiplier
Per CPU-secself-hosted models · same as Apps compute

Kimi · embeddings · rerankers · OCR · transcription · vision — every model has its own rate. See the rate sheet for the full list.

IgniteApps
$0.0432/ vCPU-hour
$0.0060/ GB-RAM-hour

Serverless containers. Pods autoscale to zero between requests.

Every plan includes a monthly allowance on every meter; on paid plans, usage past it bills per use. Need committed volume? Talk to us.

See pricing
Learn by building

Ship a real workload this afternoon.

Each tutorial is an end-to-end build on Dodil — follow it with the CLI, or hand it to your agent over MCP and watch it go.

Browse all tutorials
RAG · Vector
Build an AI research agent

Pull ArXiv papers into a persistent memory bank and answer hard questions over the embedded research.

Read the build →
Agents · Performance
A semantic cache for agents

Put an exact + near-duplicate cache in front of your agent, then track the dollars saved in a live warehouse.

Read the build →
Data · Models
A cost-gated lead pipeline

Discover companies free, classify a bounded batch on low-cost inference, then gate every paid step behind human approval — one bucket, one Ignite app.

Read the build →

Start building
in seconds.

Tell Claude Code — or any MCP agent — to build you a CRM. It picks up the skill, writes the code, and deploys with the dodil CLI. Prefer to drive it yourself? The same commands are one CLI away. Free tier, no card.

Free tier · no card required · cancel anytime
DODIL

Sovereign AI cloud. Every part of a system, one bill, wired on day one.

Sign up free
Regions
UK · LondonliveEU · FrankfurtnextMiddle EastsoonAfricasoon
Compliance
SOC 2in progressISO 27001in progressGDPR-readyData residencyenforced
© 2026 Circle Technologies Pte Ltd. All rights reserved.
PrivacyTermsSecurityAcceptable UseDPASubprocessorsBuilt for the AI era.