The Nexil platform

Knowledge agents, workflows, and the freedom to plug in any tool.

A walk through what Nexil actually does — agents, multi-step workflows, MCP integrations, Skills, models, security, and analytics.

The Challenge

Enterprise AI is stuck.

Knowledge buried in SharePoint. Workflows trapped in email chains. Every department wants its own AI chatbot, but IT can't staff the requests and security won't approve another SaaS vendor with access to company data.

✕ Disconnected tools everywhere
✕ Knowledge buried in SharePoint
✕ IT bottleneck on AI requests
✕ Security blocks new SaaS vendors
✓ Nexil: One portal. Your infrastructure.
Three ways to use Nexil

One platform, three use cases

Built on the same agent + workflow primitives — each tuned to a different job to be done. Click through for the deep dive on each.

Chatbots

Right model, right Skills, right audience

Deploy distinct chatbots to distinct teams — engineering on Claude with GitHub Skills, support on a low-cost model with the Zendesk Skill, executives on a citation-grounded assistant. One admin panel, many configurations.

See chatbots
Knowledge Agents

Answers from your documents, with sources

Purpose-built RAG agents that surface knowledge buried in SharePoint, Google Drive, Notion, Confluence, Zendesk, ServiceNow, and Sitecore. Every answer cited, every source linked, every team scoped to what it should see.

See knowledge agents
Workflows

Multi-step pipelines, no code

Turn email chains and approval forms into auditable, automated pipelines. Mix AI reasoning, REST calls, MCP tools, approval gates, and human input — every run tracked, every approval logged.

See workflows

The rest of this page covers the platform capabilities all three use cases share — MCP integrations, Skills, models, security, analytics, and the production-readiness baked into the runtime.

MCP Integrations

Connect to anything. Talk to everything.

Nexil is a first-class Model Context Protocol client. Register any MCP server you can reach over HTTPS — public, partner-hosted, or one your team built internally — with a few clicks. The examples below are a curated starter pack to get you moving on day one, not a walled garden.

Stripe

Billing & subscription tools

HubSpot

Marketing & sales automation

Linear

Engineering issue tracker

Atlassian

Jira + Confluence (single server)

GitHub

Code, PRs, issues

Notion

Knowledge base & databases

Slack

Send messages, read channels

Intercom

Conversational support

Asana

Tasks & projects

Sentry

Error tracking & releases

PostHog

Product analytics

Cloudflare

DNS, Workers, deployments

DocuSign

E-signature workflows

Brave Search

Web search for grounding

+ Bring your own

Register any public, partner, or in-house MCP server — available on every tier

Configure once at the tenant level — on every tier, register any Streamable-HTTP or SSE MCP server you can reach over HTTPS (public, partner-hosted, or one your team built and deployed) and pick which tools your workflows are allowed to call with allow-lists or block-lists.

The MCP Connections catalog with vertical-aware Recommended for you tagging

The catalog is vertical-aware. An engineering firm sees GitHub, Linear, Atlassian, Sentry, PostHog, and Cloudflare pinned at the top under Recommended for you. A legal firm gets DocuSign and SharePoint. The full catalog is always one search away — we just sort what's most relevant first.

Skills

Capability packs your agents put on like a uniform

A Skill bundles the vocabulary, output conventions, and pre-wired tools an agent needs to be useful in a specific domain — and snaps onto any agent or workflow AI step in one click. Because skills are authored in Anthropic's portable SKILL.md format, you can drop in any skill from the wider ecosystem or write your own internal ones. The examples below are a head start, not a limit.

Salesforce

Opportunities, accounts, idiomatic SOQL

Stripe

Customers, subscriptions, refund flows

HubSpot

Marketing, sales, deal pipeline reasoning

Zendesk

Ticket triage, help-center, customer context

Atlassian

Jira sprints + Confluence release notes

Linear

Engineering issues, status, project rollups

GitHub

PRs, issues, releases, commit etiquette

Slack

Channel-aware messaging & summaries

Microsoft 365

Mail, calendar, contacts via Graph

SharePoint

Site exploration & page authoring

Autodesk APS

BIM models, sheets, design data

Web Research

Brave search + page fetch with citations

SQL Analytics

Read-only SQL, schema-aware answers

Why Skills matter

Stop re-authoring prompts for every agent

Skills replace ad-hoc system prompts and one-off tool wiring with reusable, versioned capability packs. Every agent that needs to "speak Salesforce" or "act as a Stripe operator" inherits the same vocabulary and the same tool schema — instantly.

  • Two-tier registry — a Kalion-curated platform catalog plus tenant-scoped skills you author or clone-and-customise
  • Markdown-portable — standard SKILL.md format means anything written for the wider ecosystem drops in
  • Attach anywhere — pin a skill to a chat agent or to a workflow AI step; tools become callable inside the same conversation
  • Tools dispatch automatically — each skill's tool entries route to MCP servers or REST/email actions; no glue code to write
  • Auditable — every skill tool invocation is logged to per-tenant analytics for usage and cost tracking
  • Catalog as a starting point — clone a platform skill, edit the prompt or swap a tool, save it under your tenant in seconds
  • Bring your own skills — paste in a SKILL.md from anywhere, or author one for an internal system. No catalog gatekeeping, no waiting on us to add it
The Skills catalog with vertical-aware Recommended for you featuring
Production Ready

Built to actually run in production

The boring parts — timeouts, retries, retention, alerting — are baked in, not bolted on.

Step Timeouts & Retries

Per-step timeouts and configurable retry with backoff. Transient API blips don't fail your run.

Notifications

When a step needs human attention, participants get an email with a deep link — no more lost approvals.

TPM Budgets

Per-tenant tokens-per-minute caps prevent any one workflow from starving the rest. 429s with retry-after.

Cost Meter

MTD LLM spend vs tier allowance, projected end-of-month, alerts at 75/100/150% — in-portal and email.

Seat Tracking

Distinct active users tracked per billing period. Soft-warns Global Admins at 1.0× tier cap, escalates at 2.0×. Never blocks sign-in — we email instead.

Run Search & Filter

Find any run by status, date, or free-text. Debounced search across hundreds of runs.

Run Retention

Terminal-state runs older than 90 days are pruned daily — predictable storage, predictable cost.

The agent management table inside a Nexil tenant
AI Models & Self-Service Admin

Twelve models, one admin panel

Your tenant ships with twelve production LLMs hosted for you, full self-service admin, and SSO with the identity provider you already use. No Azure subscription, no API keys, no infrastructure to manage.

  • Custom subdomain at yourcompany.nexilai.app
  • SSO with Microsoft Entra ID or Clerk — MFA and conditional access inherited from your IDP
  • Full self-service admin panel for agents, workflows, MCP connections, skills, and analytics
  • Twelve production AI models included — GPT-4o, GPT-4.1, Claude, Llama, Grok, Cohere, Mistral — or bring your own LLM (Azure OpenAI, OpenAI direct, Anthropic; free toggle on Business, default on Dedicated)
Analytics & Insights

See how your teams use AI — and what it costs

Built-in usage analytics track conversations, feedback, model spend, and active-user counts across every agent. Know what's working, what each call costs, and how close you are to your tier's limits.

  • LLM cost meter — month-to-date spend vs allowance, projected end-of-month, color-coded alerts at 75/100/150% — in-portal and email
  • Active-user meter — distinct sign-ins this billing period vs your tier's seat cap, with the user list right there
  • Per-model breakdown — which agents use GPT-4o, which use Claude, which use Llama, and what each generated
  • MCP tool usage — counts and error rates per server, per tool, per tenant
  • Conversation volume + feedback — see which agents are sticky and which need work
The Nexil usage analytics dashboard with LLM cost meter and seat tracker
Security & Architecture

Zero-trust by design

Tenant-isolated data, role-based access, and enterprise-grade encryption. Nothing is shared across customers.

SSO with Your IDP

Sign in with Microsoft Entra ID or Clerk. MFA, conditional access, and session policies inherited from your IDP — Clerk in turn supports Google, Okta, GitHub, and email/password.

Role-Based Access

Group membership controls agent visibility, admin access, and workflow permissions. No Nexil-specific user database.

Tenant Isolation

Dedicated search indexes per tenant. Cosmos DB partition isolation. Blob storage scoped by tenant. No cross-tenant data leakage.

Tenant-Level Access Controls

You decide who from your company can use it

Connecting your IDP doesn't auto-enroll everyone. Map a group (Entra security group, or a Clerk organization/role) to Tenant Access and only those people can sign in. Map another to Tenant Administrators and they get Global Admin in your Nexil portal.

  • Group or user-level allow lists — resolves real names from your IDP directory
  • Self-administration — your tenant admins manage agents, workflows, MCP connections, skills, and these settings without ever calling support
  • Lock-out guard — you can never accidentally remove your own admin access
  • Audit-logged Kalion support — if you opt in, our engineers can open a 2-hour, audit-logged session as Global Admin to debug. Toggle off anytime.
  • Per-agent ACLs — plus the agent-level access lists you've always had, for tighter scoping inside a tenant
The Nexil tenant settings page showing access ACL and admin lists
Getting Started

From signup to live agent in 30 minutes

Self-service signup, free 14-day trial, no credit card. All you need is an identity provider — Entra ID or Clerk.

Sign Up & Pick Your Vertical

Pick a subdomain, choose your industry (Engineering, Legal, Healthcare, Financial Services, Education, Government, and seven more), choose Trial, click Create. We provision your portal at yourcompany.nexilai.app and pre-seed it with sample agents tailored to your vertical — skills already cloned, prompts already written.

Connect Your Identity

Pick your IDP at signup. Entra ID: your IT admin clicks one consent link — no Azure subscription, no app registration, no API keys. Clerk: sign in immediately with Google, Okta, GitHub, email/password, or any Clerk-supported source — no Microsoft tenant required.

Quickstart

Drop a single PDF, DOCX, or Markdown file. We chunk it, embed it, and spin up an instant Q&A agent — ready to chat within seconds.

Grow

Connect SharePoint, add MCP tools, build workflows from templates (or generate one with AI). Convert to a paid plan when you're ready — all your data carries over.

What's next

Plans, pricing, and a team that can build it for you

See pricing

Four plans. Free 14-day trial, $99 Starter, $999 Business (most popular), $5,000 Dedicated stamp. Per-user LLM credit consistent at $3/user.

View plans →

Implementation Services

Optional $1,000 Quickstart, fixed-price Workflow Build Packs from $4,995, and monthly Workflow Operations retainers. Money-back if we're late.

View services →

Ready to deploy your AI portal?

All you need is an identity provider — Entra ID or Clerk. No Azure subscription, no API keys, no infrastructure. 14-day free trial — no credit card required.

Start Free Trial [email protected]