Sécurité Eden AI

Know who spent what on AI, down to the call

Tag every request with the client, project, team or employee it belongs to. Charge clients back with detailed statements, give each team a budget it cannot overrun, and decide which models each team is allowed to call. One organization, one invoice, full visibility.

Parler à notre équipe

Commencer à construire en toute sécurité

Certifié SOC 2 • Certifié ISO 27001 • Prêt pour le RGPD

Monitoring · Usage by client

Platform

Models

Monitoring

API keys

Organization

Members

Guardrails

One organization, many clients, every token accounted for

Members, keys and tags give you four levels of attribution without creating a separate account per client.

Organization

ACME Agency

One credit balance, one invoice, unlimited seats

SSOSCIMAudit log

Teams & roles

Manager · Billing · Member

Plus custom roles and guardrails per role

JMKOAS

+21

API keys

One per client or project

Each with a budget, an expiry and default tags

acme-prodglobex-bot

Tags on requests

client · project · team · member

Up to 10 key/value pairs per call

client=acmeenv=prod

Sécurité certifiée

Every token attributed to a client, a project, a team or a person

Eden AI records the cost of every request together with the API key, the member and the tags behind it, so you can split your AI spend the way your business is organized.

Tags on every call

Add key/value labels to any request with the X-EdenAI-Tags header. Up to ten tags per call, counted once the header and the body are merged, on every endpoint, LLM or expert model.

Request header

X-EdenAI-Tags: client=acme,project=invoices

Recorded with the call

client=acmeproject=invoicesenv=prod

Default tags on API keys

Set default tags when you create a key and every call made with it inherits them. One key per client means zero code changes for your developers.

API key	Default tags	Budget
acme-prod	client=acme env=prod	$500 / mo
globex-bot	client=globex	$1,000 / mo
sandbox	env=test	$50 / wk

Filter on any dimension

Filter cost by tag, API key, member, model, provider or feature, and combine them to answer what the support bot cost for Acme in September.

acme

$1,912

globex

$1,240

initech

$703

Untagged spend stays visible

Whatever is not tagged lands in its own row, so you see the gap and fix the key or the call that caused it.

Untagged spend

$273

7% of total

review

Always shown as its own row, so gaps in tagging are visible immediately

One row per tag combination

The usage table carries the date, the tags and the cost in US dollars of every slice, with one column per tag key. A client, a project or an environment is one filter away.

date	tag:client	tag:project	cost_usd
2026-09-30	acme	invoices-ocr	12.18
2026-09-30	acme	support-bot	40.02
2026-09-30	globex	search	9.90

One column per tag key, dated, with the cost in US dollars

Management API

Query the same usage data over HTTP, by period, by API key and grouped by member, and feed your own billing or chargeback system.

Request

GET /v3/manage/usage/ ?start=2026-09-01 &end=2026-09-30 &api_key=acme-prod &group_by=user

Returns the same figures as the dashboard, ready for your own billing system.

Garde-fous intégrés

Charge back, budget and restrict, from one place

Tagged usage is the raw material. From there you bill your clients, cap what each team can spend, and decide which models they are allowed to call.

01 · Chargeback

Charge your clients back, with the detail to back it up

Filter a period on one client tag and you have a breakdown split by project, team or employee. Build the client-facing statement from it, while you keep one consolidated Eden AI invoice covering every provider. Talk to us about custom billing and invoicing structures.

Talk to us

Acme Corp

Draft

Usage statement · September 2026 · No. 2026-09-ACME

client=acmeenv=prod

Prepared by

Your company

Oct 1, 2026 · from Eden AI usage

Project	Requests	Cost
support-bot	812,400	$1,204.18
invoices-ocr	96,210	$388.40
search	301,870	$319.60

Total · September 2026$1,912.18

02 · Budgets

Give every team a budget it cannot overrun

Each API key carries a spending cap that resets daily, weekly or monthly. When a key reaches $0 it stops working until the next period. Guardrails add spend caps and rate limits per role, member or key, and the whole organization shares one credit balance with billing rights reserved to the roles you choose.

Read about API keys

Budgets · Teams and keys

Customer support · month to date vs. budget

Budget $2,000Spent $1,640

Customer support

82%

$1,640 / $2,000

Resets Oct 1

Data science

96%

$2,870 / $3,000

Resets Oct 1

Marketing · key paused

paused

$500 / $500

Stopped at $0 · resets Monday

Default budget for new keys$500 / month

03 · Compliance

Limit which models each team can call

Allow or deny models and providers with patterns such as anthropic/* or */*, scope rules per feature, and attach them to roles, members or API keys. Pin regulated teams to the EU endpoint, where non-EU models return HTTP 451 and never spend credits. Zero Data Retention and no training on your data by default.

Read about guardrails

Guardrail · EU-regulated teams

Allowed models

mistral/*openai/*@eu+ add pattern

Denied models

*/*

Rate limit

100 / minute

Spend cap

$500 / week

Applies to

Role · Analyst (12)API key · acme-prodMember · J. Martin

Save guardrail

Strictest rule wins

Customer stories

One contract, one bill, many providers

We are implementing a combination of sentiment analysis and topic extraction into our platform, which handles a big load of patient comments coming through a survey. It would be a hassle to setup an infrastructure to compare all the AI providers and their output, so that's why we decided for Eden AI, along with the fact that it provides us with an unified billing as opposed to maintaining an account on each of the providers.

Nikola Komes

CEO @ InsiderCX

I believe in accessing AI services through a one-stop shop that brings together major providers like Amazon and Azure, as well as emerging specialists. Eden AI is the right solution for our first ‘laboratory’ stage, giving us a portal with multiple services we can use depending on each project.

Alain Mielle

Innovation Manager @ Council of Europe

Contrôle de la gateway IA

One organization, with the roles and controls of an enterprise

Invite your whole company under one organization and one shared credit balance, then decide who can see, who can manage and who can pay.

Roles and permissions

Manager, Billing and Member roles out of the box, plus custom roles that combine organization, members and billing permissions.

Member	Role
JM Julia Martin	Manager
KO Karim Oumar	Billing
AS Ana Silva	Member
CR Chris Roy	Custom · Auditor

SSO, SCIM and audit logs

Sign in through Okta, Azure AD or WorkOS, sync your directory with SCIM and keep an administrative audit log. Contact the Eden AI team to set it up.

SSO · OktaConnected

SCIM provisioningSynced 2 min ago

Audit log

When	Who	What
14:02	J. Martin	Budget acme-prod → $1,000
13:40	K. Oumar	Guardrail EU → role Analyst
09:12	SCIM	3 members from Okta

Guardrails that stack

Reusable usage policies that set which models a team can call, how often, and how much a member can spend, attached at the level you choose.

API key

acme-prod · $500 / wk

Member

J. Martin · 100 / min

Role

Analyst · mistral/*, openai/*@eu

Org default

deny */*

Rules stack from key to member to role to organization; the strictest applicable rule wins.

Obtenir votre clé API

Lire la documentation

Comparative table

One account per client vs. tags on one organization

Separate accounts per client

Eden AI organization

One account, one invoice and one set of keys per client to maintain.

One organization and one invoice; clients, projects and teams are tags.

Spend can be split by a single dimension: the account.

Split by client, project, team, member, model or feature, combined.

Budgets live in each provider's console, if they exist at all.

Budgets per API key with daily, weekly or monthly resets, enforced at the gateway.

Model restrictions enforced by code review.

Model allow-lists and rate limits attached to roles, members or keys.

Chargeback assembled by hand from several provider exports.

Per-client breakdowns built from tagged usage, one row per tag combination.

Developer experience

Tag once, attribute everywhere

Tags travel as a header on any endpoint, or as defaults on the API key. The cost of each request comes back in the response and feeds the filters in the dashboard.

PYTHON · DEFAULT TAGS FOR A WHOLE CLIENT

# One client, one set of tags, zero changes in your call sites
from openai import OpenAI

client = OpenAI(
    base_url="https://api.edenai.run/v3",
    api_key=EDENAI_API_KEY,
    default_headers={
        "X-EdenAI-Tags": "client=acme,project=support-bot,team=cx"
    },
)

# Or set default tags on the API key itself in the dashboard:
# every call made with that key inherits them.

CURL · TAGS ON A SINGLE CALL

# Tag a single call, on any endpoint
curl https://api.edenai.run/v3/chat/completions \
  -H "Authorization: Bearer $EDENAI_API_KEY" \
  -H "X-EdenAI-Tags: client=acme,project=invoices,env=prod" \
  -d '{"model": "openai/gpt-5.4-nano", "messages": [...]}'

# Response
{
  "usage": { "total_tokens": 508 },
  "cost": 0.000202
}

# Then, in Monitoring: filter by tag and read the cost
# per client, per project or per environment.

Questions fréquentes

Obtenir votre clé API

Lire la documentation

Eden AI est-il certifié SOC 2 ?

Eden AI est-il certifié ISO 27001 ?

Eden AI prend-il en charge le Zero Data Retention ?

Eden AI utilise-t-il les données client pour entraîner des modèles d’IA ?

Eden AI est-il conforme au RGPD ?

Les données sont-elles chiffrées ?

Puis-je contrôler les providers et modèles utilisés par mon équipe ?

Eden AI propose-t-il des endpoints régionalisés ?

Eden AI peut-il router les requêtes vers des régions spécifiques côté provider ?

Que se passe-t-il si une région demandée n’est pas disponible pour un provider ou un modèle ?

Eden AI fournit-il des guardrails ?

Quelle est la différence entre Eden AI et l’utilisation directe des providers ?

COMMENCEZ

Construisez avec l’IA en gardant le contrôle et la sécurité

Attribute every token, charge back with confidence, and keep each team inside its budget and its approved models.

Parler à notre équipe

Commencer à construire

ROLES & PERMISSIONS • BUDGETS PER KEY • ONE INVOICE FOR EVERY PROVIDER

Client	Projects	Requests	Cost	Share of total
acme	3	812,400	$1,912.18	46%
globex	2	540,112	$1,240.02	30%
initech	4	301,870	$702.90	17%
umbrella	1	9,830	$0.00	BYOK
Untagged	—	64,201	$273.30	7% · review