Pricing
AI Cost Governance Pricing
Privacy-First. No Gateway. No Prompt Storage.
Control AI spend without gateways, prompt logging, or added latency. Zero prompt storage at every tier — by architecture, not policy.
💳 No credit card required
to get started ·
Annual pricing available —
contact us
Pricing in USD · Local currency billing available on request
What every plan includes — and why it matters
These aren't premium features behind a paywall. They are architectural guarantees that ship with every DoCoreAI installation.
Zero Prompt Storage
Your prompts never leave your Python environment. DoCoreAI collects only cost and token metadata — model name, token count, latency, spend. No prompt content. No response content. No PII. This isn't a setting you toggle on. It's how the SDK is built.
→ Your compliance team can approve
deployment on day one.
Predictive Budget Control
DoCoreAI trains a local LightGBM model on your actual usage patterns and predicts exactly how many tokens each request needs — before the call is made. It sets a tight, per-request token ceiling instead of the provider's bloated default. You pay for what you need, not the safety margin.
→ Eliminates 40–70% of AI token waste — your bill drops automatically.
No Gateway Architecture
DoCoreAI runs beside your application in the same Python process. It never sits in the request path between your app and the LLM provider. No proxy. No added network hop. No single point of failure. No third party seeing your traffic.
→ Zero latency impact. Zero data egress.
Nothing new to route through.
Why AI costs spiral
You pay for the ceiling.
Not the response.
Most developers set max_tokens once and forget it.
That default becomes silent waste — charged on every single API call,
across every environment, forever. Here is what that looks like in practice:
| What you're building | Default max_tokens set | Tokens actually needed | Wasted per call |
|---|---|---|---|
| Classification / Extraction | 1,024 tokens | ~300 tokens | ~70% |
| General Chat | 1,024 tokens | ~400 tokens | ~61% |
| Code Generation | 4,096 tokens | ~1,500 tokens | ~63% |
| Agentic / Long-form Tasks | 8,192 tokens | ~2,000 tokens | ~76% |
DoCoreAI trains a local model on your actual usage and predicts
exactly how many tokens each request needs — before the call
is made. It sets a tight ceiling automatically.
No code changes. No manual tuning. No guessing.
💡 Typically 40–70% of that ceiling cost disappears.
On every call. Automatically.
What the governance stack does
Core features ship with every plan. Here's what each one actually does for your AI workload.
Privacy & Security
Zero Prompt Storage
No prompt content, response content,
or PII ever leaves your environment.
Only cost metadata is collected — by
architecture, not configuration.
PII Detection at Edge
Scans outgoing prompts for personally
identifiable information before the
LLM API call is made — catching
sensitive data at the source, not
after it's been logged somewhere.
Auto-Patch LLM SDKs
Installs via pip as a monkey-patch
across OpenAI, Anthropic, Gemini,
Groq, Bedrock, and Ollama SDKs.
Zero code changes in your application.
Cost Governance
Token Prediction Engine
A local LightGBM model trained on
your actual usage predicts the exact
tokens each request needs — then sets
a tight per-request ceiling instead of
the provider's bloated default.
Budget Pacing Engine
Distributes your daily dollar budget
across 24 hours so a burst of traffic
at 9 AM doesn't exhaust the cap by
noon. Autonomous — no manual
intervention needed.
Budget Tracking & Alerts
Real-time per-request cost tracking
by team, feature, and model. Alerts
when spend approaches daily caps
or when anomalies are detected.
Scale & Reliability
Auto-Retraining
The prediction model automatically
retrains as your usage patterns shift.
No manual model management — the
governance adapts to your workload
over time.
Drift Detection
Monitors whether the prediction model
is still accurate against your live
traffic. When usage patterns diverge,
drift detection flags it and triggers
retraining.
A/B Model Testing
Compare cost and quality across LLM
providers side-by-side. See whether
switching from GPT-4 to Claude or
Gemini changes your spend profile
before committing.
Per-Team Cost Attribution
Break down AI spend by team, feature,
or environment. Know which team or
product feature drives the bill —
not just the total.
Choose your plan
Every plan includes the full governance stack. Plans differ only by volume, installations, and support.
Not sure which plan fits?
See the FAQ below ↓
·
Ask us directly ↗
Free
$
0
For teams starting their governance journey
1,000 prompts/day · 60-day pilot
Free Evaluation ↗
Run a structured 60-day governance pilot.
Self-serve setup with default configurations. By day 14, your prediction model is trained on real usage. By day 30, you have your first token waste report. By day 60, you know exactly what governance saves your team — with data to justify the upgrade.
Plan specifics
1,000 prompts/day
1 installation license
60-day governance pilot
Default app configuration — self-serve
Work email required
Community support
Startup
Best Value
$
49
/mo
For AI teams shipping their first production workloads
2,000 prompts/day · 3 licenses
Get Started ↗
Stop AI token waste from day one.
Predictive budgets enforce tight token ceilings across your production workloads — eliminating 40–70% of AI token waste. Before you go live, a short questionnaire configures governance precisely for your use case. Your optimal settings arrive by email. No manual tuning.
Plan specifics
2,000 prompts/day / license
2 installation licenses
Governance configured for your workload
Email support
Team
$
199
/mo
For teams governing AI across multiple products
5,000 prompts/day · 10 licenses
Get Started ↗
Govern AI costs across your full product suite.
10 installation licences mean dev, staging, and production across multiple products — all governed independently. Budget pacing, token prediction, and per-team cost attribution run across every environment. Configured for your workload before day one.
Plan specifics
5,000 prompts/day / license
5 installation licenses
Governance configured for your workload
Email support
Scale
$
499
/mo
For platforms with high-volume, multi-environment AI
8,000 prompts/day · 20 licenses
Get Started ↗
High-volume AI governance at platform scale.
20 installation licences across your full platform stack. Peak-aware pacing handles traffic spikes without blowing your monthly cap. A/B model testing compares provider costs on real traffic. Governance configured for your platform architecture before you go live.
Plan specifics
8,000 prompts/day / license
8 installation licenses
Governance configured for your workload
Email support
Enterprise
$
1499
/mo
For regulated enterprises
15,000 prompts/day · 30 licenses
Book an Intro Call ↗
Govern AI costs across teams and business units.
Reduce organisational risk with governance configured precisely for your compliance boundaries, team structure, and workload before deployment. Starts with a founder onboarding call within 24 hours.
Plan specifics
15,000 prompts/day / license
15 installation licenses
Founder onboarding call within 24 hours
Dedicated Customer Success Manager
Priority support
Annual billing · Invoice & PO accepted
Built for your exact requirements.
Volume beyond Enterprise, specific compliance boundaries, or unique deployment constraints?
We'll build the right configuration — prompts, licenses, retention, and integrations — around your needs.
Build your plan
Custom prompts/day volume
Custom installation count
Custom metadata retention
Custom integrations
Volume discounts
Annual billing available
30k+
PyPI downloads
20+
Enterprise engagements
6+
LLM providers supported
17mo
Active development
🛡️ Security by Architecture
Your prompts never touch our servers
DoCoreAI executes inside your application runtime. Your prompts, customer payloads, and API keys never touch our servers. There is nothing to retain — and nothing to breach. The only data that reaches the DoCoreAI cloud is aggregated metadata: model name, token count, latency, and cost. That's it.
🔍 PII Detection — What gets blocked
DoCoreAI scans every outgoing prompt for email addresses, SSNs, credit card numbers, phone numbers, and IP addresses — before the LLM call is made. If a match is found, the call is hard blocked. Not redacted. Not passed through with a warning. Blocked entirely, in-process. An alert is written to the developer console. No PII value is stored anywhere.
⚠️ Technical prompts containing IP addresses (e.g. log analysis, network debugging) may trigger the PII filter. This is by design — configure exclusions for known-safe patterns via the governance config.
🔑 API Key Handling
DoCoreAI monkey-patches your LLM SDK at the Python process level. It intercepts calls after your SDK has handled authentication — it never reads, stores, or transmits your API keys. Your keys stay inside your SDK client object, in your runtime, under your control.
📋 Compliance Readiness
GDPR
✅ Supported
Zero prompt storage eliminates personal data processing at the LLM layer.
HIPAA
🔄 Architecture Ready
No PHI stored or transmitted — formal certification in progress.
Get started in minutes
Three commands.
Your code stays untouched.
No SDK wrappers. No refactoring. No changes to your existing LLM calls.
3 commands — run once per environment
# Install
$ pip install docoreai
# Configure — token from docoreai.com + settings from onboarding email
$ docoreai config
# Start governance — run once, stays active
$ docoreai start
$ pip install docoreai
# Configure — token from docoreai.com + settings from onboarding email
$ docoreai config
# Start governance — run once, stays active
$ docoreai start
Your existing code — completely unchanged
# No imports. No wrappers. Nothing added.
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user",
"content": "Summarise this report."}]
# max_tokens set automatically by DoCoreAI
)
DoCoreAI intercepts the call, predicts the
exact tokens needed, sets a tight ceiling —
and lets it through.
Your code never knew it was there.
Every call from this point is governed,
tracked, and cost-optimised automatically.
Frequently asked questions
Are my prompts stored at any tier?
+
Never, at any tier.
Zero prompt storage is a core
architectural principle of DoCoreAI —
not a paid feature. Only cost, token
counts, latency, and model name are
collected. No prompt content, no
response content, no PII ever leaves
your environment.
What is a "prompt" in billing terms?
+
One prompt equals one call made to
an LLM provider — one completion, one
embedding, one generation. Retries count
as separate prompts. Batch calls count
as one prompt per item in the batch.
How does the 60-day Free evaluation
work?
+
You get full platform access for 60
days — the complete governance stack,
every feature, up to 1,000 prompts per
day. That's enough time to train the
prediction model on your real data, see
your first cost report, and evaluate
whether DoCoreAI fits your workflow.
After 60 days, choose a paid plan to
continue. A work email address is
required to register (no gmail, hotmail,
or similar). Your metadata is retained
for 3 months after the evaluation
expires.
What does a "License" or "installation" mean?
+
An installation/license is one Python environment where DoCoreAI is installed and active — typically one each for development, staging, and production. Free includes 1 installation license. Startup includes 3, Team includes 10, Scale includes 20, and Enterprise includes 30. Custom plans can include any number of installations.
What happens if I exceed my daily
prompt limit?
+
Prompts above your daily limit are still
tracked locally, but telemetry will not
sync to the cloud dashboard until the
next day. Core budget control and pacing
continue to work locally regardless.
Upgrade to a higher plan for
uninterrupted cloud sync.
Can I upgrade from Free to a paid plan?
+
Yes. Your local telemetry data, trained
prediction models, and configuration
carry over when you upgrade. There is
no migration required — just connect
your account to a paid plan.
Why does DoCoreAI not use a gateway?
+
Gateways sit in your request path — every
LLM call routes through a third-party
proxy. That adds latency, creates a
single point of failure, and sends your
prompt data through an external system.
DoCoreAI runs inside your application
process via SDK monkey-patching. Your
app calls the LLM directly, unchanged.
DoCoreAI listens locally, extracts
metadata, and applies governance —
without ever touching the request path.
How does predictive token budgeting
reduce costs?
+
Most applications set
max_tokens to 2,000+ as
a safety net, but most responses need
a fraction of that. You pay for the
ceiling on every request. DoCoreAI
trains a LightGBM model on your actual
usage and predicts the exact tokens
needed per request — then sets a tight,
per-request ceiling. Over thousands of
calls, this eliminates silent token waste
and reduces spend 40–70%.
Is DoCoreAI production-ready?
+
The core SDK — budget control, pacing,
governance, prediction, and PII detection
— is production-ready at v2.1.0,
built on 17 months of research and
validated across 20+ enterprise
AI engagements. The cloud dashboard
enters public beta in July 2026.
Do you offer annual pricing?
+
Annual pricing with a discount is
available. Contact us directly:
[email protected]
Is a Data Processing Agreement
(DPA) available?
+
Yes. A DPA is available on request for
all paid plans. Given DoCoreAI's
zero-prompt-storage architecture, the
only data processed by our servers is
aggregated metadata — model name, token
count, latency, and cost. No prompt
content, no user data, no PII ever
leaves your environment. Contact us to
receive the DPA template:
[email protected]
Where is my metadata stored?
+
All metadata synced to the DoCoreAI
cloud dashboard is stored on
infrastructure hosted in Manchester,
England (United Kingdom) via Hostinger.
The UK operates under UK GDPR —
substantially equivalent to EU GDPR.
Local telemetry (your SQLite database)
never leaves your own environment.
For teams requiring specific data
residency outside the UK, contact us
to discuss deployment options:
[email protected]
What happens if DoCoreAI fails
to initialise?
+
DoCoreAI activates via a
⚠️ Known issue — work in progress. A startup health check that explicitly confirms governance is active is on our development roadmap. Until then, always validate your installation in a staging environment before production deployment by confirming DoCoreAI appears in
.pth file injected into
your Python environment at install time.
If the .pth file fails to
load — due to a misconfigured Python
path, version mismatch, or container
build issue — Python silently skips it.
Your application continues running
normally but with zero
governance applied. LLM calls
go through unmanaged and untracked,
with no error or warning raised.
⚠️ Known issue — work in progress. A startup health check that explicitly confirms governance is active is on our development roadmap. Until then, always validate your installation in a staging environment before production deployment by confirming DoCoreAI appears in
pip list and checking
that your Python environment matches
the installation target.
Does DoCoreAI work in Docker
and Kubernetes?
+
Docker deployments are supported with
one requirement: DoCoreAI must be
installed into the same Python
environment that runs your application
inside the container — not mounted as
a volume or installed on the host only.
Add
⚠️ Kubernetes — not yet validated. Multi-pod and orchestrated Kubernetes deployments have not been formally tested at this stage. If you are deploying to Kubernetes and want to be an early tester, contact us directly — [email protected] — we will work through it with you.
pip install docoreai
to your Dockerfile after your other
dependencies. The .pth
activation mechanism works identically
inside containers as it does in
standard Python environments.
⚠️ Kubernetes — not yet validated. Multi-pod and orchestrated Kubernetes deployments have not been formally tested at this stage. If you are deploying to Kubernetes and want to be an early tester, contact us directly — [email protected] — we will work through it with you.
For Procurement & Legal
Everything your team needs to approve DoCoreAI
All vendor documentation available on request. Enterprise deals supported with full procurement paperwork.
📄 Data Processing Agreement
Available on request for all paid plans.
Given zero prompt storage architecture,
only aggregated metadata is processed
on DoCoreAI servers.
Request DPA →
Request DPA →
⚡ SLA & Uptime
Formal SLAs available on
Enterprise plan —
dedicated infrastructure and defined
uptime guarantees targeted
September 2026.
Discuss SLA requirements →
Discuss SLA requirements →
💳 Payment Methods
All plans:
Credit card (monthly billing)
Enterprise only: Invoice and purchase order supported. Wire transfer available on request.
Enterprise only: Invoice and purchase order supported. Wire transfer available on request.
📋 Contract Terms
Startup, Team, Scale:
Month-to-month. Cancel anytime.
No lock-in.
Enterprise: Annual commitment available. Volume discounts on multi-year terms.
Enterprise: Annual commitment available. Volume discounts on multi-year terms.
🌍 Data Residency
Cloud metadata stored in
Manchester, England (UK)
via Hostinger. UK GDPR compliant —
substantially equivalent to EU GDPR.
Local telemetry never leaves your own environment.
Local telemetry never leaves your own environment.
🛡️ Security Documentation
Security architecture publicly
documented at
docoreai.com/security/
Vendor security assessment and additional documentation available on request for Enterprise. SOC 2 audit targeted Q3 2026.
Request security docs →
Vendor security assessment and additional documentation available on request for Enterprise. SOC 2 audit targeted Q3 2026.
Request security docs →
Need something not listed here?
Email us directly
— procurement requests are handled
by the founding team, not a ticketing system.
Not sure which plan is right? Let's talk.
Book a 20-minute call with Saji — founder and CEO. We'll understand your use case, compliance requirements, and usage patterns, and recommend the right starting point. No sales pressure.