Back to Home

AI at scale,
under control.

Budget controls, usage tracking, AI model policies, access rules and enterprise self-hosted deployment, everything IT and finance need to roll out Saga with confidence.

See all controls
Saga Admin, Governance Overview
Admin
Monthly spend
$3,240
of $5,000 limit
Active users
148
across 6 teams
Model policies
12
active rules
Company token budget, January
64.8%
3.24M / 5M tokens used12 days remaining
Top users this month
JS
James S.
Engineering
440k
MK
Maria K.
Product
312k
DL
Dana L.
Sales
218k
What Governance & Compliance covers

Built for the teams who say yes or no

Budget controls
Set hard spending limits at the company, team or individual level. Soft warnings at 80%, hard stops at 100%, no surprise invoices.
Usage tracking
Real-time dashboards showing who uses what, how much, and when, broken down by user, team, feature and time period.
AI model policies
Decide which models each user group can access, give power users GPT-4o, restrict contractors to lighter models, block experimental ones entirely.
Self-hosted deployment
Run Saga entirely within your own infrastructure, VPC, on-prem or air-gapped. Your data never leaves your perimeter.
Policy rules
Define what AI can and cannot do company-wide, block certain data sources, restrict model access by role, enforce prompt-level guardrails.
SSO & SCIM
Native integration with Okta, Azure AD, Google Workspace and any SAML 2.0 provider. Provision and deprovision users automatically via SCIM.
Alerts & reports
Automated alerts for budget thresholds, unusual activity spikes and policy violations, delivered to Slack or email.
Access controls
Role-based access across every Saga feature, admins, editors, viewers and custom roles, tied to your existing directory groups.
Feature 01

Spend limits that actually hold

AI costs can spiral fast without guardrails. Saga's budget controls let you set spending limits at every level, company-wide, per team, per feature, or per individual, with automated warnings before you hit the ceiling and hard stops that never let you go over.

Monthly token budgets at company, department, team and user level
Soft alerts at configurable thresholds (e.g. 70%, 90%), Slack or email
Hard stops at 100%, graceful degradation, not silent failure
Budget carry-over and rollover rules configurable per team
Finance-ready cost export, CSV, PDF or direct integration with your billing tools
Budget usage, January 2025
Engineering
58%
Sales
83%
Marketing
101%
Product
41%
Company monthly limit
$5,000Edit
Marketing exceeded their $800/mo budget. Usage has been paused. Increase limit or wait for next cycle.
Feature 02

See exactly where AI is going

Usage tracking gives you a live, granular picture of how Saga is being used across your organisation, by person, team, feature and time. Know who's getting value, catch unusual spikes early and make informed decisions about where to invest or cut back.

Real-time dashboard, token consumption, request volume, active users by team
Breakdown by feature, Chat, Meetings, Agents, Digests, API calls
Per-user drill-down, see individual usage history and session logs
Daily / weekly / monthly trend charts with period-over-period comparison
Anomaly detection, automatic alerts on unusual spikes or off-hours activity
Usage, last 7 days
Day
Week
Month
Mon
Tue
Wed
Thu
Fri
Sat
Today
Breakdown by feature
Chat
1.24M
48%
Meetings
720k
28%
Agents
390k
15%
Digests & API
234k
9%
Feature 03

The right model for the right person

Not every user needs access to every model, and not every model costs the same. Saga's model policy engine lets admins decide exactly which AI models each user, group or team can access. Give power users the full range, restrict contractors to lighter options, and keep experimental models off by default.

Assign model access at the individual, group or team level, or set a company-wide default
Supports all major model families, GPT-4o, Claude 3, Gemini, Llama and custom private models
Block experimental or high-cost models from general users with one toggle
Model access tied to cost, budget limits and model tiers work together automatically
Self-hosted users can connect private models and restrict them to specific groups only
Model access
Available models
GPT-4o
GPT-4o mini
Claude 3.5 Sonnet
Gemini 1.5 Pro
o1 Preview
Custom-internal
Engineering & Product
52 users
GPT-4o
Claude 3.5 Sonnet
Gemini 1.5 Pro
GPT-4o mini
Sales & Marketing
38 users
GPT-4o mini
Gemini 1.5 Pro
Contractors & External
14 users
GPT-4o mini only
Feature 04

Your infrastructure, your rules

For organisations where data cannot leave their perimeter, Saga ships as a fully self-contained deployment. Run it in your AWS, Azure or GCP VPC, in your own data centre, or completely air-gapped, with the same feature set as the cloud version, none of the external calls.

Helm chart + Terraform modules for Kubernetes deployments on any cloud
Air-gapped mode, no outbound internet required after initial install
Bring your own LLM, connect to your private Azure OpenAI, AWS Bedrock or on-prem model
Managed updates via a dedicated release channel, your team controls when to upgrade
Dedicated customer success engineer included for all self-hosted Enterprise plans
Deployment configuration
SagaAI Cloud (SaaS)
Fully managed. Hosted on GCP with EU or US region selection. Zero ops overhead.
Self-Hosted (VPC / On-Prem) Selected
Run inside your own perimeter. AWS, Azure, GCP or bare-metal Kubernetes supported.
Air-Gapped
No outbound traffic. For classified environments and regulated industries.
Runtime
Kubernetes
LLM provider
Azure OpenAI
Data region
EU-West-1
Feature 05

AI that operates within your policies

Every organisation has guardrails for what AI should and shouldn't do. Saga's policy engine lets admins codify those rules, block access to specific data sources, restrict which models certain roles can use, enforce PII redaction, and require human approval before agents take certain actions.

Source-level blocks, exclude HR, legal or board-level data from AI access by default
Role-based model access, contractors and interns can use lighter models only
PII auto-redaction, names, emails and identifiers masked before leaving your perimeter
Agent action approvals, require human sign-off before any destructive or external action
All policy changes are versioned, roll back to any previous state in one click
Policy rules
PII auto-redaction
Names, emails and phone numbers are masked in AI responses for non-admin users.
Block HR & Legal from AI access
Confluence HR space and Legal Slack channels excluded from all AI queries.
Agent action approval
All agent write actions require manager approval before execution.
Restrict contractors to GPT-4o mini
External users limited to cost-efficient models only.
Enterprise Sales

Ready to deploy Saga on your terms?

Talk to our enterprise team about self-hosted deployment, model policies, custom contracts and volume pricing.

View Pricing

Self-hosted available · SSO & SCIM · Custom SLAs · Volume pricing