Compare·Model aggregation gateway
Continuum
Multi-agent workbench
VS
OpenRouter
Model aggregation gateway

Continuum vs OpenRouter

Metered access to every model versus a flat weekly allowance fused to the workbench that spends it.

OpenRouter is the default answer when an application needs many models behind one OpenAI-compatible endpoint. Its homepage advertises 500+ models, 80+ providers, 200T+ monthly tokens, and 10M+ global users, with the pitch stated plainly as better prices, better uptime, no subscriptions. You buy credits, you spend them per token, and the fee structure is published: 5.5% on Stripe credit purchases with a $0.80 minimum, 5% on crypto purchases, and 5% on bring-your-own-key traffic above a $25,000 monthly allowance on pay-as-you-go. Continuum is a different product with an overlapping bill. It is the free workbench where a developer runs Claude Code, Codex, Cursor, Grok, and OpenCode under subscriptions they already own, each managed session in its own git worktree and branch, with plan-approval gates, per-hunk diff review, and PR create, review, and merge in the same session. Its hosted inference is optional and flat: Plus at $25, Max 100 at $100, Max 200 at $200, and Ultra at $500 per month, each with a weekly hosted-usage allowance of $25, $100, $200, or $1,000. The split is clean. OpenRouter wins when the job is arbitrary model access for an application. Continuum wins when the job is coding agents and you want the meter to stop being a per-token decision.

Download for Mac All comparisons
Workflow Monday path Matrix Decide Money FAQ

Updated 2026-08-03 · Mac stable · Win/Linux desktop beta

Choose Continuum when

Pick Continuum when the workload is coding agents and you want worktree isolation, plan gates, diff and PR review, live quota gauges, per-repo cost, and a flat weekly hosted allowance instead of a per-token decision on every turn.

Choose OpenRouter when

Pick OpenRouter when you need the widest model catalog behind one endpoint, per-token credits with no subscription, programmatic key provisioning with per-key spend caps, and routing for an application rather than for an agent workbench.

Snapshot direct alternative
Dimension Continuum OpenRouter
Primary job Run and supervise coding agents Access many models from one endpoint
Model reach Provider agents plus optional hosted router 500+ models across 80+ providers
Billing shape Free app · flat monthly hosted tiers Prepaid credits · per token · no subscription
Stated fees No fee on BYOK subscription sessions 5.5% Stripe · 5% crypto · 5% BYOK over $25k/mo
Git isolation Worktree + branch per session Not a repository tool
Review loop Plan gate · diff · PR create/review/merge No review surface
Spend controls Weekly caps · model policy · per-repo ledger Per-key credit limits with scheduled resets
Mac stable · web · iPhone · Watch · Win/Linux desktop beta · free app
01

Metered aggregation vs coding workbench

In Continuum

A week of agent work on a fixed allowance

Mon 09:20

Start Claude Code and Codex in separate worktrees on the same repo, each on its own branch, under the subscriptions already on the machine.

Tue 14:05

Approve a plan from the phone, interrupt the weaker run, and let the better one continue without opening the laptop.

Thu 11:30

Review the diff hunk by hunk, open the pull request from the session, and watch checks without leaving the transcript.

Fri 16:45

Read cost by repo for the week and see that two repositories account for most of it. The hosted allowance did not move.

In OpenRouter

A week of model access for an application

Mon 09:20

Point the app's base URL at OpenRouter and try four models behind the same request shape without signing four contracts.

Tue 14:05

Mint a scoped key through the management API with a credit limit and a monthly reset, one per customer instance.

Thu 11:30

Restrict routing to providers whose training and retention policy the compliance review will accept.

Fri 16:45

Reconcile the invoice: token spend plus the 5.5% credit purchase fee, and decide whether BYOK is worth the 5% above the allowance.

02

Monday path

Monday with a metered gateway and a flat allowance side by side

Install, connect, first session - steps you can run the same day.

01

Write down whether the tokens are being spent by an application you are building or by coding agents you are supervising. That answer decides most of this page.

02

If it is an application, point it at OpenRouter, buy a small credit balance, and read the fee line on the receipt: 5.5% with a $0.80 minimum on a Stripe purchase, 5% on crypto.

03

Provision a scoped OpenRouter key through the management API with a credit limit and a daily or monthly reset, so one runaway loop cannot drain the balance.

04

If it is coding agents, install Continuum and start two worktree sessions on one bounded ticket under the subscriptions you already pay for. That path costs nothing.

05

Put one session in plan mode, approve the plan, review the diff hunk by hunk, and open the pull request from the same session.

06

Add an OpenRouter key inside Continuum if you want its catalog in the same workbench. The two layers stack; they do not have to be chosen between.

07

Only then compare bills: a week of OpenRouter credits burned against a Continuum hosted tier whose weekly allowance is fixed at $25, $100, $200, or $1,000.

08

Check Continuum's cost by repo, provider, model, and day before committing to either. Most teams discover the spend is concentrated in two repositories.

How Continuum prices and gates agent work

How Continuum makes agent spend legible instead of per-token

  • Every managed session runs the provider's real agent under the login or subscription the developer already owns, so BYOK work has no gateway fee at all.
  • Each session receives its own git worktree and branch before any edit lands, which is what makes running several agents at once safe rather than merely possible.
  • Plan mode is read-only and approval-gated. Approval respawns the same session with write permission and carries the accepted plan forward.
  • Live quota gauges read the provider's own rate-limit windows, so a batch is not launched into a window that is about to close.
  • Local agent logs are priced into tokens and dollars by repo, provider, model, and day without routing anything through a proxy.
  • Optional hosted inference is OpenAI-compatible at https://continuumcode.ai/v1 and Anthropic-compatible at the bare origin, with cont_sk_ keys and an auto model router.
  • Org controls cover model policy, weekly spend caps, and member approvals, so the flat allowance has a governance surface rather than just a number.
03

Gateway and workbench matrix

Cells use product nouns on both sides. Continuum is the multi-provider workbench; they win where their loop is the product.

Capability Continuum OpenRouter
Finished coding-agent workbench Mac, iPhone, Watch, web, Windows and Linux beta An API endpoint and dashboard, not a workbench
Run Claude Code, Codex, Cursor, Grok, OpenCode First-class provider sessions under your subscriptions Serves models to any client that speaks its API
Model catalog breadth Provider agents plus optional hosted auto router 500+ models, 80+ providers per the homepage
Per-session git worktree and branch Isolation before the first edit lands No repository layer
Plan approval, diff review, PR merge Plan gate, per-hunk review, PR in session Out of scope for a gateway
Works with subscriptions you already pay for BYOK-first; no fee on those sessions BYOK supported; 5% above a $25,000 monthly allowance
No per-token decision on each turn Flat tiers with weekly allowances of $25 to $1,000 Credits are metered per token by design
Published fee schedule Four monthly tiers with named allowances 5.5% Stripe with $0.80 minimum, 5% crypto
Free usage without paying The whole app is free with your own subscriptions Free models capped at 50 requests per day, 1,000 with $10 in credits
Programmatic key provisioning with spend caps Org weekly spend caps and model policy, not per-key minting Management API keys with credit limits and daily, weekly, or monthly resets
Live provider quota gauges Reads each provider's own rate-limit window Credit balance and usage tracking, not subscription windows
Cost by repo, provider, model, and day Local-first ledger with no proxy required Usage reporting is per key and per model, not per repository
Data policy controls BYOK sessions stay on the provider you chose Account and per-request routing limited to providers by training and retention policy
04

Where the two bills diverge

01 · Unit

One sells tokens, the other sells a place to spend them.

OpenRouter's product is the endpoint: 500+ models, 80+ providers, one request shape, credits that drain per token. Continuum's product is the session: worktree, branch, plan gate, diff, pull request, quota gauge, and a cost ledger. The overlap is only the optional hosted inference tier, and which model either one spends on is a question our model directory answers with one card per model.

02 · Bill shape

Metered credits versus a flat weekly allowance.

OpenRouter is explicit that there is no subscription, and its fees are published as percentages: 5.5% on a Stripe credit purchase with a $0.80 minimum, 5% on crypto, 5% on BYOK traffic above $25,000 per month. Continuum hosted inference is $25, $100, $200, or $500 per month with a weekly allowance of $25, $100, $200, or $1,000. Neither is cheaper in general. Metered wins on bursty low volume; flat wins when an agent loop makes per-token math a daily distraction.

03 · Coding agents

Agent traffic is not the same as app traffic.

A coding agent reads files, retries, and runs long turns, so its token curve is spiky and hard to predict from a feature spec. That is exactly the shape that makes per-token billing feel like a tax on iterating. It is also the shape that makes worktree isolation and a plan gate matter, and a gateway has neither.

04 · Governance

Per-key caps versus per-repo truth.

OpenRouter's controls are strong at the key boundary: credit limits, scheduled resets, a BYOK inclusion toggle, and usage tracking per key. Continuum's controls are at the human and repository boundary: model policy, weekly spend caps, member approvals, and spend attributed to a repo, provider, model, and day. A platform team wants the first. An engineering manager asking which project burned the budget wants the second.

05 · Stack, not swap

OpenRouter runs inside Continuum.

OpenRouter is a supported provider in Continuum through the OpenCode connector, so its catalog can appear in the same workbench as the subscription agents. Treating this page as a strict either-or would be dishonest. The real question is only which one pays for the tokens.

05

Coexistence

Stack recipe

How people run both.

These two are not mutually exclusive, and the honest version says so. OpenRouter is a supported provider inside Continuum: it rides the OpenCode connector, so an OpenRouter API key configured in Continuum puts OpenRouter models in the workbench alongside the subscription agents. The competition is only between OpenRouter credits and Continuum hosted inference as the thing that pays for tokens. The workbench layer above them is not contested.

Download Continuum Providers
06

Pick by what the tokens are for

Continuum

Choose Continuum for coding agents you supervise

  • The tokens are being spent by coding agents you supervise, not by an application you are shipping.
  • You want several agents running at once in isolated worktrees without them fighting over one checkout.
  • Plan approval, per-hunk diff review, and in-session pull requests are part of the job, not a separate tool.
  • A fixed weekly allowance is worth more to you than the lowest possible per-token price.
  • You need spend attributed by repository and provider, and weekly caps that an org admin sets.
  • You already pay for Claude, Codex, Cursor, or Grok subscriptions and want them used first.
OpenRouter

Choose OpenRouter for broad model access in an app

  • Your application needs many models behind one OpenAI-compatible endpoint.
  • Breadth matters more than depth: 500+ models across 80+ providers is the feature.
  • You want prepaid credits with no subscription and no monthly commitment.
  • You mint keys programmatically per customer instance and need per-key credit limits with scheduled resets.
  • Routing must be restricted to providers whose training and retention policy passes your review.
  • Volume is high enough that BYOK matters, and the $25,000 monthly fee-free allowance covers you.
07

What you pay

Continuum

App + your labs

$0 for the app on Mac, iPhone, Watch, web, and the Windows and Linux desktop beta, running under the provider subscriptions you already pay for. Optional hosted inference is $25 per month for Plus, $100 for Max 100, $200 for Max 200, and $500 for Ultra, with weekly hosted-usage allowances of $25, $100, $200, and $1,000.

OpenRouter

Their bill

No subscription. You buy credits and spend them per token at each model's rate. Published fees: 5.5% on Stripe credit purchases with a $0.80 minimum, 5% on cryptocurrency purchases, and 5% on bring-your-own-key usage above a $25,000 monthly allowance on pay-as-you-go, or above $200,000 on Enterprise. Free models are limited to 50 requests per day, rising to 1,000 per day once at least $10 in credits has been purchased.

How to compare

Total cost of work

These are different bills for different jobs. Continuum charges nothing for the workbench and nothing for BYOK sessions; its paid tiers buy a fixed weekly hosted allowance. OpenRouter charges nothing to hold an account and everything per token. Compare them only after deciding whether the spend is agent work or application traffic.

OpenRouter plans checked 3 August 2026 against their pricing page. Vendors reprice often - if you spot a stale figure, tell us and we will fix it. Continuum never marks up a session it runs under your own login.

Continuum ladder: Free app · Plus $25/mo · Max 100 · Max 200 · Ultra - full pricing. Also analytics, multi-account, devices.

08

Questions

Continuum vs OpenRouter.

Deep dives: docs, providers, sessions.

Only for the inference half. Continuum hosted inference competes with OpenRouter credits as a way to pay for tokens, at $25, $100, $200, or $500 per month with weekly allowances of $25, $100, $200, and $1,000. Continuum does not replace OpenRouter's catalog of 500+ models for an arbitrary application, and OpenRouter does not replace Continuum's worktrees, plan gates, diff review, or pull requests.

Yes. OpenRouter is a supported provider in Continuum through the OpenCode connector, so an OpenRouter API key puts its models in the same workbench as your Claude Code, Codex, Cursor, and Grok sessions. The two products stack.

Model tokens at each provider's rate, plus published platform fees: 5.5% on Stripe credit purchases with a $0.80 minimum, 5% on crypto purchases, and 5% on bring-your-own-key usage above a $25,000 monthly allowance on pay-as-you-go. There is no subscription.

It has free models rather than a free allowance of paid ones. Without purchased credits, free models are capped at 50 requests per day in total; with at least $10 in credits purchased, that cap rises to 1,000 requests per day.

If you already hold a Claude subscription, Continuum runs Claude Code under that login with no gateway fee, and adds worktree isolation, a plan gate, and diff and PR review around it. OpenRouter is the better answer when you need models that your subscription does not include, or when the client is an application rather than an agent.

It exposes retention as a routing filter rather than a single switch. You can set an account-level policy on whether providers that may train on your data are allowed, separately for paid and free models, restrict individual requests to providers with a given data policy, and filter to providers with zero-retention policies. Enterprise customers can additionally keep processing inside an EU or US region. Verify the current provider table before designing a compliance story around it.

Begin

Run your agents
in Continuum.

Free app. Your subscriptions. Optional hosted inference. Mac stable - Windows and Linux desktop are beta.

Download for Mac See pricing

vendor-neutral · local-first · multi-device