Internal unified API
Expose OpenAI Chat, Responses, and Anthropic Messages. Cursor, Claude Code, and internal apps all point at the company gateway.
Private AI gateway for your company.
Spend, conversations, privacy —
under your control.
1Provider runs inside your network. Employees, IDEs, and internal services use the same API. Your vendor contracts and keys stay unchanged. Spend, full conversation records, and sensitive-data review stay under your control.
VPC / data center / private cloud · single binary · data stays in-house
PrivateRuns in your environment
Full textEvery request and session is auditable
ScanSearch and review sensitive patterns across all sessions
GovernRoles, quotas, usage governance
01
A control plane for company AI traffic
One ingress, routing, audit, and privacy. Not a token marketplace. Not a panel that only forwards calls.
Protocol translation
Client and upstream protocols can differ. Translate among four OpenAI, Anthropic, and Gemini formats while keeping one internal standard across vendors.
Virtual models
Employees only see names you define. Exact match, prefix wildcards, or a catch-all — opaque to callers.
Failover
Round-robin across keys. On timeout, 429, or 5xx, fail over automatically, then probe again after a 300-second cooldown. Applications do not notice.
Request & session text
Keep the full text of every call. Rebuild complete conversations so leaders can see what employees asked the models.
Sensitive-data review
Scan every historical session with eight built-in pattern types plus custom regular expressions. Find the employee, session, and exact context behind each match.
Usage and cost
Roll up tokens, latency, and estimated cost by employee, role, and virtual model. Configure monthly budgets and a five-hour rolling rate limit.
Team console
Dashboard, policies, providers, logs, sessions, and RBAC with 23 granular permissions. Operations and security share one console.
02
Record every employee-AI interaction
Every request and every session is a record the company can review.
Request-level text
- Every HTTP status code and request header
- The original JSON structure of every request
- Fast troubleshooting
Session-level dialogue
- Keep every session record
- Replay the complete employee-AI exchange
- Analyze it with the built-in agent
By person
Search by employee, API key, or role. Security teams and managers see how AI is used.
Assisted review
A built-in analyst can query requests, sessions, and sensitive patterns — review by asking, not grepping logs.
03
Scan sensitive data across every session, then review by asking
Keys, identity documents, and internal identities: the built-in analysis agent scans every conversation for patterns, connecting discovery, location, and traceability.
Eight built-in pattern types
API keys, AWS keys, PEM private keys, password fields, Authorization headers, national IDs, phone numbers, and email addresses. Extend them with custom RE2 expressions.
Scan the full session database
The built-in agent searches historical sessions for sensitive patterns and returns truncated context for every match. See the employee, session, and exact passage immediately.
Read-only views + SQL audit trail
Analysis only uses read-only views with sensitive columns removed, and every query is logged. Original sessions remain intact for traceability without uncontrolled access.
04
Keep the model vendors your company already uses
Use your own contracts and keys instead of buying through another public relay. Keys are stored with AES-GCM encryption.
●OpenAI●Anthropic●Gemini
●DeepSeek●Kimi●Any compatible endpoint
05
Four private steps to take back AI traffic
Deploy in-house
Run in your VPC, data center, or private cloud. Single binary, SQLite. No public relay in the path.
Attach existing vendors
Use the Base URLs and keys from contracts you already have. The relationship stays yours.
Name internal models
coding, cheap, audit-safe. Round-robin or fallback — the gateway enforces it.
Point everyone at it
Claude Code, Cursor, Codex, internal services. After the Base URL change, audit and privacy policy apply to all.
06
One internal API. Any upstream dialect.
Employees keep their SDKs. The gateway translates, routes, records, and checks every request.
OpenAI Chat
POST /v1/chat/completionsOpenAI Responses
POST /v1/responsesAnthropic Messages
POST /v1/messagesModel list
GET /v1/models
07
Delivered as infrastructure, not a take-rate on tokens
Private install
Go backend embeds the console in one binary. No extra cluster required. IT can ship it under existing rules.
Data stays local
Config, request text, sessions, and token stats live in your database. Back up one file, take the state with you.
Roles and access
Admins, security teams, and department leads see different scopes of sessions and usage. Assign 23 granular permissions instead of sharing one admin key.
Existing clients
Built-in guides for Claude Code, Codex, and more. Change env vars, switch the company to the internal gateway.
08
How 1Provider differs from OpenRouter
OpenRouter is a public model marketplace for everyone. 1Provider is a private gateway for a company. Both can look like “one API, many models.” They are not the same product.
Public relay
OpenRouter-style services
- Requests go to their cloud, so prompts leave your company boundary
- You buy models through their account; the vendor contracts are not yours
- Built for individual developers and the public, with usage-based billing
- You cannot govern internal AI use by employee or session
Private
1Provider
- Deploy in your VPC, data center, or private cloud; traffic stays in your environment
- Keep the OpenAI, Anthropic, and domestic model contracts you already purchased
- Built for companies and teams: one ingress, permissions, quotas, and audit
- Every employee-AI conversation is visible and searchable
Put company AI traffic behind your own gateway
Private deployment. Your vendors. Every conversation auditable. Sensitive data searchable across the full session database.
See how private rollout works