See every AI coding tool your team uses — before you try to control it
Models, cost, and device health across your org. Self-hosted and open source.
| Developer | Cursor | Claude Code | Copilot | Ollama |
|---|---|---|---|---|
sarah@acme.dev Platform | $48.20 | 1.2M tok | 890ms | — |
alex@acme.dev Backend | $36.10 | 840K tok | 1.1s | — |
morgan@acme.dev Infra personal key detected | $29.80 | 620K tok | 1.4s | — |
jordan@acme.dev Mobile | $18.40 | 410K tok | 760ms | 2 models |
taylor@acme.dev Data | $12.10 | 280K tok | 2.1s | running |
Supported tools & runtimes
The problem
Your team already runs a dozen AI coding tools. Your organization cannot see any of it.
Developers pick Cursor, Claude Code, Copilot, or local Ollama models on their own machines. Engineering leaders have no org-wide view of adoption, spend, latency, or misconfiguration. Finance sees surprise invoices. Security cannot tell personal keys from company keys.
Observability answers for engineering, finance, and security
The MVP delivers the foundation: who uses what, what it costs, how it performs, and whether devices are configured correctly.
24 active devs
Who is using AI coding tools
Per-developer adoption across your org — active users, last seen, and coverage gaps.
8 tools tracked
Which tools & models
Tool installations and model calls broken down by user, team, and device.
$284.50 / 24h
Cost by user / tool / model
Estimated spend attribution so finance and eng leads see where money goes.
1.2s avg · 0.8% err
Latency & failure rate
Per-model response times and error rates to spot degraded providers early.
31/33 online
Device configuration health
Agent heartbeats, config drift, and unhealthy device flags in one view.
3 personal keys
Personal vs company keys
Detect when developers use personal API keys instead of org-provisioned ones.
12 local runtimes
Local models on devices
See which developers run Ollama, LM Studio, or other local inference.
12,847 req / 24h
Request activity log
Inspect individual requests with model, tokens, latency, and status.
How it works
Four steps to org-wide visibility
Observation requires no change to the developer's workflow. Install, enroll, configure, see.
Install the admin app
Deploy the self-hostable UseJunction admin on your infrastructure — Docker Compose, a single VM, or your existing cluster. Your data stays on your network.
Roll out the local agent
Developers enroll devices with a single token. The lightweight agent runs locally and reports metadata — no workflow changes required.
Configure supported tools
Point Cursor, Claude Code, Continue, and other supported tools at the agent. Observation only — no traffic interception.
See usage in the dashboard
Cost, models, latency, device health, and key usage appear in one org-wide view. Visibility before control.
Open source
Your data never leaves your infrastructure
UseJunction is open source under the MIT license. Self-host on a single machine, Docker Compose, or your own cloud — no vendor SaaS holding your usage data.
Pricing
Simple, transparent pricing
Open source and self-hosted. No hidden fees — your data stays on your infrastructure.
Community
Perfect for side projects and small teams
per month
- Self-hosted MIT license
- Up to 25 enrolled devices
- Usage & cost dashboard
- Tool & model visibility
- Community support on GitHub
Team
For growing teams that need more power
per month
- Everything in Community
- Unlimited enrolled devices
- Per-developer cost attribution
- Latency & error rate metrics
- Personal key detection
- Device configuration health
- Team API keys (roadmap)
- Multi-team rollup (roadmap)
- Policy controls (roadmap)
Enterprise
For organizations that need full control
contact us
- Everything in Team
- Dedicated onboarding
- Custom retention policies
- Deployment assistance
- Priority support channel
- SSO & SAML (roadmap)
- Audit log export
- SLA guarantee
Where this is going
Visibility today. Control next.
Once your organization can see AI coding usage, UseJunction becomes the layer where teams introduce team API keys, adopt local models, and eventually add routing, policy, and cost optimization. Roadmap — not available today.
Incoming requests
UseJunction · router
routing by cost · latency · policy
Routed to
—Claude Sonnet
Anthropic
GPT-4o
OpenAI
Ollama
local
Copilot
GitHub
Route requests to the right model based on cost, latency, and policy — when you're ready.
Roadmap — not available todayTeam API keys
Introduce org-provisioned keys once usage data shows where personal keys hide.
Local & self-hosted models
Help teams adopt Ollama and open-weight models with visibility into local usage.
Policy & cost optimization
Enforce budgets and model policies on top of the observability foundation you already trust.
Start with visibility. Grow into control.
Deploy the open-source admin app, roll out the agent, and see what your team actually uses.