Which tools can an agent use?

Any MCP server: register one of the 1,000+ public servers or your own, attach it to an agent, and bind credentials at the scope you choose, the whole organization, one agent or one user. OAuth-backed servers can authenticate a person mid-chat. The tools read your systems in place, scoped to the caller, with no copy of the data.

How does authentication work?

Two ways. Per user: each person signs in with the identity they already have, every call carries their own token, and roles are enforced on each request. Server to server: a tenant-scoped platform API key, sent as an Api-Token header, authenticates your backend and can mint per-user tokens. The platform key lives on your server, never in browser code.

How is it billed, and how do I control cost?

By usage, never by seat, on one bill across every model and tool. Spend caps exist for the whole organization, for one agent, and for one person on one agent, each with a reset period and a mode that blocks or alerts; enforcement is server-side. The Cost tab shows spend by provider, model, user and individual call.

📅 Book a 30-min Demo 📞 Call/text (571) 293-0242

Agentic Router

One API key for every model and tool

Q: What is Agentic Router?

Agentic Router is the engineering door to the ibl.ai platform: one API key that reaches every model the platform serves, every tool connected over MCP, and every agent your organization has built, from REST, WebSocket, the SDK, agent skills (Claude Code, Codex, Cursor and others, documented at ibl.ai/developer/skills) or any MCP client. One bill, spend caps at three scopes, and with ibl.ai you own all the code and the data.

Q: Which models can I call?

On the organization recorded on November 6, 2026, the catalogue listed 45 providers and 316 chat models: Anthropic, Google, OpenAI, Meta, Mistral, DeepSeek, Qwen, xAI and the open-weight labs, through OpenRouter, Vertex, Foundry, Bedrock and each provider's own API. The list syncs itself every six hours, and a new model goes live only after a verification gate has tested recall, streaming, tool calling, image input and a real cost.

Q: Where does it run?

In your cloud, on-premise or fully air-gapped, where the same grid points at models served inside the perimeter and your own provider keys are preferred over platform keys. Or hosted by ibl.ai to start. Either way the catalogue, the keys, the routing rules and the agents sit in your own database.

One API key for a firm's legal-ops and IT developers: every model, 1,000+ MCP servers and the firm's agents, callable from the matter system or the client portal, per attorney or server to server, with one bill and spend caps. On ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing — so you can deploy anywhere, from your own cloud to a fully air-gapped network. No need to choose build vs. buy — you get both.

Sector:Higher Education Corporate Small Business K-12 Government Legal Financial Services Healthcare

1.6M+Platform Users

400+Organizations

“ibl.ai does amazing work and are great people.”
Paul LeBlanc, former President and CEO, Southern New Hampshire University

Don’t take our word for it?See what our partners say →

We host itStart at $0 on the ibl.ai cloudFree tier with renewable monthly credits, no credit card, usage-based from there. Move to your own infrastructure anytime, nothing to rebuild.Hosted deployment See pricing

You own itTake the whole codebaseBackend and infrastructure under a perpetual license, so you own all the code and the data. Your cloud, on-premise or air-gapped, under your control.Full code license

Available on

What is behind the key

Every model the platform serves, every tool reachable over MCP, and every agent your organization has built, from REST, WebSocket, the SDK, agent skills or any MCP client, on one bill with ceilings that hold.

Every model, one endpoint45 providers and 316 chat models on one grid, through OpenRouter, Vertex, Foundry, Bedrock and each provider's own API. Switch a model per request or per agent; prompts, data and connectors stay. The catalogue syncs itself behind a verification gate.45 providers · 316 models Every tool, over MCPRegister any of the 1,000+ public MCP servers or your own, attach it to an agent, and bind credentials for the organization, one agent or one user. Tools read your systems in place; no copy of the data.1,000+ MCP servers Your agents, two ways to authenticateCall any agent you built from your own code with the user's own token and role, or server to server with a tenant-scoped platform API key. One bill, spend caps at three scopes, cost per call.REST · WebSocket · SDK · skills · MCP

How is Agentic Router different from a model gateway?

A model gateway puts one key in front of several model vendors and bills you per token. Agentic Router puts one key in front of the models, the tools and the agents you own, authenticates people and services differently, and runs where your data already is.

	Agentic Router	A hosted model gateway
What the key reaches	Agentic RouterEvery model, every MCP tool, and every agent you built.	A hosted model gatewayModels only.
Authentication	Agentic RouterPer-user tokens with roles enforced, or a tenant-scoped platform key for services.	A hosted model gatewayOne organization key for everyone.
Billing and caps	Agentic RouterOne bill across models and tools; caps per organization, agent and user; cost per call.	A hosted model gatewayPer-model spend, one cap if any.
Keys and routing	Agentic RouterYour provider keys first; the highest-priority gateway that can serve the model.	A hosted model gatewayThe vendor's keys, the vendor's routing.
Where it runs	Agentic RouterYour cloud, on-premise or air-gapped, with models served inside the perimeter.	A hosted model gatewayThe vendor's cloud; every prompt passes through it.
What you own	Agentic RouterThe code and the data.	A hosted model gatewayA subscription.

Your Code. Your Data.Full ownership, zero vendor lock-in

Any LLM, Your ChoiceAir-gapped or any cloud provider’s

85% Lower CostSignificantly cheaper at scale

Engineers IncludedOur team builds it with yours

MCP Architecture Guide for Legal

How to connect your legal systems (document management, case management, billing, research) to AI agents — step by step.

Read the Guide

Models & Runtimes — any model, any agent runtime

Google Gemini Microsoft Copilot Amazon Bedrock OpenAI GPT Anthropic Claude OpenClaw NemoClaw

Services

AI Transformation Air-Gapped AI Studio Production

Agentic Router — One API Key for Every Model, Every Tool and Every Agent You Own

Agentic Router by ibl.ai gives a firm's developers one key to everything the platform serves: 45 providers and 316 chat models, the MCP tools that reach Clio, iManage and DocuSign in place, and the agents the firm has already built.

Call it from the matter system, intake or the client portal, with each attorney's own token or a platform key, with privilege-aware roles enforced on every call.

Deployed in your infrastructure with your keys and your controls—including full code access.

For The firm's legal-ops and IT developers: One Key, Every Model and Tool

Every model, one endpoint45 providers and 316 chat models on one grid as of November 6, 2026: Anthropic, Google, OpenAI, Meta, Mistral, DeepSeek, Qwen, xAI and the open-weight labs, through OpenRouter, Vertex, Foundry, Bedrock and each provider's own API. Switch a model per request or per agent; prompts, data and connectors stay.

Every tool, over MCPRegister any of the 1,000+ public MCP servers or your own, attach it to an agent, and bind credentials at the scope you choose: the whole organization, one agent or one user. The tools reach Clio, iManage and DocuSign in place, with no copy of the data.

Your own agents, from your own codeEvery agent your organization has built on Agentic OS is callable from the matter system, intake and the client portal: a REST call, a streaming WebSocket session, the SDK, an agent skill or any MCP client. For example: "Run the conflicts-check agent from the intake form, server to server, before a matter opens."

ResultOne integration instead of a vendor account per model and a connector per tool.

For Platform Teams: One Bill, Enforced Ceilings

Unified billingEvery model and every tool behind the key lands on one bill, priced by usage, never by seat. The Cost tab breaks spend down by provider, model, user and individual call, with latency percentiles beside it.

Spend caps at three scopesA ceiling for the whole organization, for one agent, and for one person on one agent, each with its own reset period and a mode that blocks or alerts at the limit. Enforcement is server-side, so a runaway script stops at the cap.

Your keys firstA request routes to the highest-priority gateway that can serve the model and has a usable key, and your organization's own provider keys are preferred over platform keys. A deployment with provider keys and no ibl.ai key works with no extra configuration.

A catalogue that keeps itself currentThe model list syncs every six hours and a new model goes live only after a verification gate has tested multi-turn recall, streaming, tool calling, image input and a real, non-zero cost. For example: "Send first-pass contract review to one model and a client-facing summary to another, from one key."

ResultCost you can see per call and cap per scope, with no procurement cycle per model.

For Security & IT: Two Ways to Authenticate, One Place to Audit

Per userAttorneys, staff and clients sign in with the identity they already have, and every call carries that person's own token. Roles are enforced on each request, so an agent answers only what the caller may see, and a person's in-chat model choice stays inside the list you allow.

Server to serverA tenant-scoped platform API key, sent as an Api-Token header, authenticates your backend: mint per-user tokens, provision users, run pipelines. It can act for any user on the tenant, so it lives on your server, never in browser code.

Where it runsYour cloud, on-premise or fully air-gapped, where the same grid points at models served inside the perimeter. For example: "Cap each practice group's monthly spend and see cost per matter."

Every call loggedWho asked, which model answered, which tools were called and what it cost, on systems you administer, with replayable transcripts and an audit trail per agent.

ResultOne key policy for people and for services, and nothing to reconcile across vendors.

How Do You Call Agentic Router?

RESTMore than 1,000 documented endpoints: agents, sessions, datasets, memory, MCP servers and connections, users, roles, analytics and billing. Authorization: Api-Token <PLATFORM_API_KEY> for a service, or the user's own token.

StreamingChat sessions stream over WebSocket, so a response renders token by token in your own interface.

SDK and CLI@iblai/iblai-js for web and native apps, @iblai/cli for the terminal, and a Python quickstart (iblai/quickstarts) that creates an agent and chats with it from a script.

Agent skills161 skills from iblai/vibe teach Claude Code, Codex, Cursor, OpenCode or any skills-compatible coding agent to run the platform itself: configure agents, manage datasets and memory, pull cost analytics, each mapped to exact REST calls. Every skill is documented at ibl.ai/developer/skills; install with npx -y @iblai/mcp.

Any MCP clientEvery agent you deploy can be added to Claude Code, Cursor, Claude Desktop or any other MCP client in one step, and talked to from there with its own knowledge, memory and guardrails.

Embed and proxyEmbed the chat widget in your own app with tokens provisioned from your backend, no login popup. Third-party services such as ElevenLabs and HeyGen sit behind the same key through the external service proxy.

What's Inside Agentic Router (Highlights)

Model catalogue45 providers, 316 chat models, synced every six hours behind a verification gate

GatewaysOpenRouter, Vertex, Foundry, Bedrock and each provider's own API, your keys preferred

MCP servers1,000+ public servers plus your own; credentials bound per organization, agent or user; OAuth-backed connectors

Your agentsEvery agent on Agentic OS callable by ID, with its knowledge, memory and guardrails

Two auth modesPer-user tokens with roles enforced, or a tenant-scoped platform API key for services

Spend capsOrganization, agent and per-user ceilings; block or alert; reset periods

Cost per callSpend by provider, model, user and call, with latency percentiles

SurfacesREST, WebSocket, @iblai/iblai-js, @iblai/cli, 161 agent skills for Claude Code, Codex and Cursor, any MCP client, embed, external service proxy

Why Agentic Router Instead of a Model Gateway?

What the key reachesAgentic Router: Every model, every MCP tool and every agent you own | Model gateway: Models only

AuthenticationAgentic Router: Per-user tokens with roles, or a tenant-scoped platform key | Model gateway: One organization key for everyone

BillingAgentic Router: One bill across models and tools, caps per organization, agent and user | Model gateway: Per-model spend, one cap if any

Where it runsAgentic Router: Your cloud, on-premise or air-gapped, your provider keys first | Model gateway: The vendor's cloud, through the vendor's keys

Your dataAgentic Router: Read in place over MCP, scoped to the caller, every call logged on your systems | Model gateway: Every prompt passes through a third party

OwnershipAgentic Router: You own all the code and the data | Model gateway: A subscription

Benefits That Drive Impact

The firm's legal-ops and IT developersOne key and one SDK instead of a vendor account per model and a connector per tool.

Platform teamsOne bill, cost per call, and ceilings that hold without a meeting.

Security and leadershipTwo auth modes under one policy, every call logged, and the whole stack owned rather than rented.

Part of the ibl.ai platform

Agentic Router is how engineers reach the platform: Agentic OS runs the agents, Agentic Ontology exposes your systems over MCP, Agentic Vibe builds the apps that call them, and Agentic LMS, SIS, Wallet and CRM are applications built the same way. One key, one bill, one platform you own.

Experience Agentic Router Today

Start for FreeStart at ibl.ai/join, open an agent in Agentic OS, and create a platform API key on its API tab

Read the skills161 agent skills documented at ibl.ai/developer/skills, one per capability, each mapped to exact REST calls, for Claude Code, Codex, Cursor and any skills-compatible agent

Read the API referenceEvery endpoint, live from the platform's OpenAPI document, at ibl.ai/developer/api-reference

Schedule a DemoWalk through your integration, your models and your cost controls with our engineers

What our partners say about us

Case Studies

Chris Gabriel | Google

Case Studies

Lorena Barba | George Washington University

Case Studies

Dr. Juana Mendenhall | Morehouse College

Juile Diop | MIT

Adam Tetelman | Nvidia

Jason Dom | American Public University System

Benjamin Breyer | Columbia University

Ken Fujiuchi | SUNY

Erika Digirolamo | Monroe College

David Flaten | SUNY

David Vise | Modern States Education Alliance

Linda Wood | ARM Institute (U.S. Department of Defense)

Case Studies

Chris Gabriel | Google

Case Studies

Lorena Barba | George Washington University

Case Studies

Dr. Juana Mendenhall | Morehouse College

Juile Diop | MIT

Adam Tetelman | Nvidia

Jason Dom | American Public University System

Benjamin Breyer | Columbia University

Ken Fujiuchi | SUNY

Erika Digirolamo | Monroe College

David Flaten | SUNY

David Vise | Modern States Education Alliance

Linda Wood | ARM Institute (U.S. Department of Defense)

Frequently Asked Questions

Bring us a hard problem. Keep the code.

Run Agentic Router on your own terms

Start free on our cloud today, or take the whole codebase into your own infrastructure. Either way our engineers work alongside your team, and you own all the code and the data.

Get Started Free Book a 30-minute demo Read the skills

ibl.ai is family-owned and operated from New York, NY — a U.S.-headquartered, domestically-owned long-term partner, not a vendor that sells licenses and moves on. · SOC 2 Type II certified