What your people can do
99 features across 22 parts of the product, each one making your people faster, your costs clearer or your company safer. New ones land every week.
Interfaces
Web app
Anyone can ask any approved model, from the browser, with nothing to install.
Model picker
Every model your company allows, grouped by lab, with Auto first.
Price before you send
See what a message will cost on each model before you send it.
Cost on every answer
Every reply shows the model, who picked it and what it cost.
Grounded in your code
Answers cite your repositories, Wiki and connected tools.
Sessions
Every chat is kept, searchable and private until you share it.
Team switcher
Ask as yourself or on a team’s budget and model rules.
Interfaces
Desktop app
Engineers run coding agents in their own repos, safely.
Claude Code and Codex
Run both in your own repos, through your gateway.
Ask first, then run
Agents ask before they run commands or open pull requests.
Same budgets as the web
One budget and one audit trail across every surface.
Local and cloud
Start locally, hand off to a Cloud Agent when you close the lid.
Interfaces
Mac widget
Ask, approve and keep an eye on agents without leaving your work.
Quick ask
Press ⌥ Space and ask any model from anywhere.
Approvals
Approve an agent’s command or pull request in one click.
Running agents
See what is running, on which model, at what cost.
Team spend
Your budget and your team’s, at a glance.
Interfaces
Mobile apps
Approve and stay informed from anywhere.
Notifications
Agents ask for approval; budgets warn before they run out.
Approve with Face ID
Approve a command or a pull request from your lock screen.
Artifacts on the go
Shared answers open natively on phones and tablets.
Continue any session
Pick up a chat you started on the web or in the terminal.
Interfaces
Slack agent
Teams get every AI where they already talk.
@Nexus in any channel
Ask in a channel or a DM; the answer comes in a thread.
Auto per step
Auto picks the cheapest model that can do each step.
Name a model
Say “with Claude Opus 5.5” to pick the model yourself.
Fixes as pull requests
Drafts the fix as a PR with buttons to review it.
Interfaces
CLI and API
Engineers and agents build on every model with one endpoint.
nexus claude, nexus codex
Your coding agents on your keys and budgets.
OpenAI and Anthropic SDKs
Change the base URL; keep your code.
nexus/auto
One model name that routes every request to the right model.
Cost on every response
Model, policy, cost and budget left, in every response.
MCP server
Your gateway’s models and tools, available to any MCP client.
A command for every rule
Every admin task has a command and an endpoint.
Capabilities
Artifacts
Turn any answer into a page your team can read, reuse and share.
Charts, tables and code
Answers become clean pages people can read and reuse.
Sources and models
Every artifact shows what it was built from and by which models.
Share with anyone
Teams, people, a guest by email, or a link, under your rules.
Versions
Every share is a version you can revoke at once.
Capabilities
Wiki
Everyone understands the codebase, always up to date.
Written from your repos
Architecture, services, owners and how to run it.
Always current
Pages refresh daily or weekly as the code changes.
Ask about any page
Questions answered from the page and the code behind it.
Grounds every agent
Chats and agents read the Wiki before they answer.
Capabilities
Review
Every pull request gets a careful first review in minutes.
Blocking issues first
The comments that matter, ranked above the nits.
Fixes in one click
Suggested changes you can commit from the comment.
Your rules
Review follows your Wiki and your written standards.
Routed to owners
Each review goes to the people who own that code.
Capabilities
Cloud Agents
Hand off long work to agents that keep going without you.
Persistent workspaces
Agents keep their environment between runs.
Runs while you’re away
Close your laptop; the work keeps going.
A trace of every step
Model calls, tools and waits, step by step.
Approvals
Agents pause for you before anything that matters.
Capabilities
Synced sessions
Start anywhere and pick up anywhere.
One session, every device
Terminal, browser, desktop and phone share one session.
Private until shared
Your sessions are yours until you share them.
Platform
Enterprise controls
Give everyone AI without losing control.
One set of rules
The same checks on web, desktop, mobile, Slack, CLI and API.
Refused before it’s spent
A call over budget or to a blocked model never reaches the provider.
Audit of every call
Who asked, which model, what it cost, and why.
Data rules
Keep customer data inside the company, by policy.
Platform
Organization
Mirror how your company is organized, as deep as it goes.
Nested units
Divisions, teams, people and agents, as deep as you need.
Admins and roles
Every unit has its own admins and roles.
Rules only narrow
A team can never allow what its division blocks.
SSO and SCIM
Sign in with your identity provider; accounts stay in sync.
Platform
Budgets and allowlists
Keep AI spend predictable, team by team.
A budget per unit
Monthly budgets for every division, team and person.
Allow or block any model
Model policies per unit, inherited downward.
Limits per person
A default limit for everyone, raised where needed.
Alerts
Warnings before a budget runs out, on every device.
Platform
Connectors
Let AI work with the tools your company already runs.
Your whole stack
Code, tickets, docs, chat, CRM, data and monitoring, by their real names.
Connect once
Every model, agent and surface can use a tool after one connection.
Read-only first
Every connection starts read-only; widen it when you choose.
Scoped per team
Each team sees only the tools and data it should.
MCP servers
Bring any MCP server and it becomes a connector with the same rules.
Every call audited
Each tool call is logged with who, which agent and why.
Platform
Observability
See who uses AI, on what, and what it costs.
Spend per team
AI spend per person, team and unit, across every provider.
Traces for every call
Open any model call routed through your gateway, per team.
Download anything
Traces and audit as CSV or JSON, in the app and through the API.
Prompts, only if you allow
Prompt and response text stays off until an owner turns it on, then expires.
Adoption
Who uses AI, on which surfaces and models, team by team.
Delivery metrics
Throughput, cycle time and stalls per team, from GitHub and Linear.
Spend per change
What each merged change cost in AI, from your own ledger.
Weekly digest
A summary in Slack every week: spend, adoption and stalls.
Ask with sources
Plain-language questions about your org, answered with sources you can open.
Platform
Audit
Answer any auditor’s question in minutes.
A full audit log
Who asked, which model, which surface, what it cost and what was decided.
Policy decisions
Every allow and every refusal, with the rule that made it.
Admin changes
Every change to units, roles, budgets and model rules.
Retention you set
Keep logs as long as your policies require, then delete them.
Export
Send the log to your SIEM or download it for an auditor.
Platform
Bring your own keys
Keep the keys, plans and discounts you already have.
Your API keys
Free up to $25k a month of usage, then 5%.
Your AI plans
Claude and ChatGPT plans, used through Nexus at no fee.
Your discounts
Provider contracts and discounts stay yours.
Models
Explorer
Anyone can find the right model for the job.
Rankings by task
Every model ranked for coding, writing, reasoning and more.
Model pages
Benchmarks, price, speed, context and the lab behind it.
Linked from the picker
Compare models without leaving your chat.
Models
Auto Router
Pay less without anyone choosing models by hand.
Per-step routing
Each step goes to the cheapest model that can do it.
Inside your rules
Auto only picks models your policies allow.
Every choice explained
A reason for every routing decision.
Models
Benchmarks
Know a switch is worth it before you make it.
Score against cost
Every model plotted, with Auto on the frontier.
Your own tasks
Run the benchmark on a sample from your repositories.
Enterprise
Enterprise
Run Nexus wherever your data must live, with the support you need.
SSO and SCIM
Sign in with your identity provider (SAML, OIDC); accounts stay in sync.
Managed
We run it for you, isolated in our cloud; a dedicated region on request.
Private
Run it in your AWS, GCP or Azure account, or on your own servers.
Zero data retention
Model providers keep nothing from your requests.
Private networking
Reach your gateway over private links, never the open internet.
Complex organizations
Nested units, roles and budgets for every division, team and agent.
SOC 2 and ISO 27001 programs
Controls and evidence for your security reviews.
Your keys, your contracts
Bring your provider keys and AI plans; keep your discounts.
Dedicated account team
A team that knows your setup, plus dedicated engineers for rollout.
Highest priority support
The fastest response, from people who build the product.