one endpoint for every AI model

Use any AI model. Bring every account.

Connect the provider accounts you already pay for. DSCO bills $2–$20 every month for each account you keep connected. Or route through DSCO Managed Cloud at DSCO's metered per-model rates. Send model:"auto" and each request goes to the model that fits — cheap when the task is easy, frontier when it isn't.

Create a workspace → See how it works
MONTHLY PER CONNECTED ACCOUNT · every active account connection costs $2–$20 each month; DSCO Managed Cloud is metered separately at DSCO per-model rates that include our usage margin.
Providers
15 labs
Models
—
Every call
logged + traced
Spend
hard caps
// what you get
You keep control of every call.

Point the accounts you already pay for at one OpenAI-compatible URL. Each connected account has a clear monthly control fee; you keep the keys, pay the provider directly, and see exactly where every call routed.

01

Bring your own accounts

Every provider account you keep connected costs $2–$20 each month, based on its plan. Calls run on your credentials, and every request shows which connected account served it.

02

Automatic failover

Send model:"auto" and the router picks from the providers you allow, then fails over to the next one when a provider errors or rate-limits you.

03

No rewrite

Keep your OpenAI or Anthropic client. Point its base URL at the router, and it translates each provider's format for you.

04

Spend you can cap

Set lifetime and 30-day limits per key. Managed spend is held before a call runs, so usage can't pass the cap you set.

05

Broad model catalog

Hundreds of priced models across leading providers, searchable and reachable through the same endpoint.

06

Move your agents over

Point Claude Code, Codex, or OpenCode at the router with one reversible config change. No new client stack.

// live status
Running right now.

Pulled from this deployment in real time.

// playground · real paid route
Route one request. See exactly what happened.

The playground runs a real request against your prepaid balance, then shows the answer, route, cost, and remaining credit.

workspace / playground
01
Create a workspace
Your keys, routing settings, and usage records stay together.
02
Add credits
Choose the amount. The router checks your balance before a paid request runs.
03
Route a request
Get the real model output plus its route, usage, and cost.
Create a workspace →
✓
The actual outputNo canned answer or synthetic success screen.
✓
The route and costSee what ran, how much it used, and what it charged.
✓
The same production pathThe playground uses the same account and routing boundary as the API.
// developer setup
Switch with one base URL.

Keep your OpenAI or Anthropic client. Point it at DSCO Router, then choose models, provider accounts, and spending limits from your workspace.

—
models indexed on this node
Browse the full catalog, connect accounts, and configure routes without scrolling a raw provider dump.
Open catalog & console →
# point any OpenAI client at your router
export OPENAI_BASE_URL="https://router.distributed.systems/v1"
export OPENAI_API_KEY="sk-dsco-…"

# or wire your coding agents in one command
export DSCO_ROUTER_API_KEY="sk-dsco-…"
dsco-router configure --api-key "$DSCO_ROUTER_API_KEY"

# opt in per session; Router sessions default to GPT-5.6 Terra xhigh
# ordinary Claude/Codex defaults stay untouched
claude --settings ~/.claude/dsco-router.settings.json
codex -p dsco-router
// questions
Questions you're probably asking.

The honest answers, including the ones about price and terms.

What does it cost to connect my own accounts?
DSCO charges a recurring monthly fee of $2–$20 for every provider account you keep connected, based on the value of its provider plan. The fee continues each month while that account connection is active; it is not a one-time connection charge. It pays for encrypted credential handling, health checks, routing, failover, spend controls, and request records. Your provider continues to bill its plan or API usage directly. Any requests routed through DSCO Managed Cloud are billed separately at DSCO's metered per-model rates.
Is routing through my own accounts against my provider's terms?
Your accounts remain isolated to your workspace and your own usage. DSCO charges for controlling each connected account; it does not pool, share, or resell capacity from your accounts to anyone else. Every call stays on your credentials and provider relationship. Provider terms differ, so DSCO exposes only account types supported for the relevant provider and plan.
What does DSCO Managed Cloud cost?
It's DSCO's metered capacity across OpenAI, Anthropic, and open models, billed from prepaid credits at a DSCO rate for each model. DSCO sets each managed model's retail price, which includes our usage margin. You set hard spend caps per key, and every call shows exactly what it cost.
Do I have to rewrite my code?
No. The router is OpenAI- and Anthropic-compatible. Point your existing client's base URL at the router, keep your code, and you're routing. Coding agents like Claude Code, Codex, and OpenCode move over with one reversible config change.
What does model:"auto" actually do?
It hands the request to Chimera, which picks the model that fits the task: a cheaper one when that's enough, a frontier one when it isn't. It routes on capability, and every call shows which model ran and why. Streaming requests use Chimera's model selection and stream from that model directly.
Where do my keys and data go?
Provider keys are encrypted at rest and used only to route your own calls. Usage records show the model, cost, latency, and route reason. They do not store your prompts or responses.
Can I leave, or am I locked in?
No lock-in. Change providers, models, and fallback order from your workspace without touching your code, and disconnect any account whenever you want. It's your accounts and your keys.
Why not just use OpenRouter or Ramp Router?
Those resell tokens through one endpoint, which is genuinely useful. DSCO also lets you connect accounts you already pay for and charges a recurring monthly fee for every account you keep connected. You see the account, route, and cost on every call; DSCO Managed Cloud is the separate metered option, priced by DSCO with an included usage margin.
Who is this for?
Developers and teams who run more than one provider account and want one endpoint, spend caps, and a record of every call. If you only want the single cheapest tokens with zero setup and don't care about bringing your own accounts, a flat reseller may fit you better.