Invite codes no longer gate your own accounts. Bring-your-own-account use is free for everyone; cloud Pro stays invite-only.Read the announcement
AnnouncementAugust 9, 2026

To everyone who's been following or using Mirasim:

Over the past few days, Mirasim ended up in front of a much bigger audience much earlier than we'd planned. We were still in private beta and, honestly, nowhere near ready for a wider launch.

Invite codes spread far beyond the small beta group we had planned for. In a very short window, we saw a flood of visits and test requests, along with some unusually aggressive, probing request patterns. Our capacity, reliability, and security systems weren't ready for that volume. Some of you ran into slow loading, failed requests, or Pro benefits that never showed up.

We're genuinely sorry for the rough experience. Whatever the source of the traffic, we should have been better prepared. That's on us, and our users shouldn't have to pay for it.

At the same time, you sent us a huge amount of honest, detailed feedback. A lot of issues that probably wouldn't have surfaced until a public launch showed up early. Thank you to everyone who took the time to try Mirasim, patiently share feedback, or report security issues responsibly.

Here's how access to Mirasim will work from here

Invite codes no longer limit connecting your own accounts to Mirasim — they only limit the cloud Pro service that Mirasim pays for.

Connecting and using your own accounts locally will be free for everyone. You'll be able to connect the Claude Code, Codex, and other model accounts or services you already use, then work across multiple models, Agents, and sessions in Mirasim. Mirasim won't charge for any of this. Third-party subscription or API fees will still follow each provider's own terms.

Cloud Pro access — where Mirasim covers the model usage costs to make things cheaper for users — will remain invite-only and roll out in batches as capacity allows. We'll keep sharing our progress and what opens next, so the rules stay as clear and transparent as possible.

A note on Pro benefits

  • The 1,000 beta users covered by our original promise will still receive one full year of Mirasim Pro. That promise has not changed.
  • Anyone who successfully registered during this wave after we had already passed the original 1,000-user limit will receive one month of Mirasim Pro. This is both to make up for the unstable experience and to thank you for trying Mirasim before it was fully ready.
  • If you've already registered or activated your access but your Pro benefits still aren't showing up, please don't try again. We have the records. We'll verify eligibility and apply the benefits in batches, then share the timing and start-date details.

How the rollout will work from here

Over the next week or two, we'll focus on adding capacity, fixing the entitlement system, tightening security, and working through your feedback one issue at a time. Mirasim remains open. Registration and local, bring-your-own-account access will stay available, while cloud Pro access will roll out in batches as capacity allows.

We're not trying to build yet another model aggregator. We want Mirasim to be an Agent workspace you can keep building in for the long run. Wherever the next leap in Agent capability comes from, you should be able to plug it into the same workflow, switch seamlessly, and use different Agents together. Once you've built something, you can put it in front of simulated users in a real environment and let evidence — not guesswork — tell you what to do next.

We want every leap in what Agents can do to become real leverage for everyone — so you spend less time chasing tools and more time building what you actually care about.

Models will change. You shouldn't have to start over.

Thank you for showing up before Mirasim was fully ready. From here, we'll put these promises into every fix and every update, and work to earn your trust by making Mirasim more stable and more reliable.

— The Mirasim team

Mirasim
Get helpGet Mirasim

Mirasim Documentation

Chapter 10

Models, accounts and quota

Swap the model and reasoning level per session, bring your own keys or run local, juggle several accounts, and fall back when quota runs dry.

Model and reasoning level are per session

The composer carries three pickers — agent, model, reasoning level — and they apply to that session, not to the app. So the design pass can run on a strong reasoning model in one pane while the mechanical follow-through runs on something faster and cheaper in the next.

The model list is grouped: Native for the models your agent's own subscription serves, My Models for profiles you configured yourself. Reasoning level is offered where the model actually has one; where it does not, the picker says so rather than showing a control that does nothing.

The three pickers, and the account behind them
The composer's model and reasoning-level pickers, and the account and quota indicator in the title bar.

ModelPer session. Native subscription models and your own profiles appear together, grouped — and a long-context variant is marked [1m]. — this control is not on the captured screen.

The buttons in this panel work — click them to look around.

Which agent to run

Agent, model and tool are configured independently. Adding a provider does not require a new workbench, and adding a tool does not couple it to one agent.

The agents Mirasim drives
Mirasim's built-in agent
Ships with the app, no separate install. Runs in-process and reaches models through whichever route you configured, including the relay.
Claude Code
Anthropic's coding CLI, reliable in long, tool-intensive sessions. Its own sign-in and its own terminal UI are preserved exactly.
Codex
OpenAI's terminal coding agent, strong at planning and multi-step refactors. Also runs with its own auth and UI intact.
Pi GUI
Not a coding agent — this one works a browser. It reads the page the way a screen reader does, from the structure the browser exposes for assistive software, rather than guessing from screenshots. Its models run through the relay, so it needs no key of your own. A coding agent can hand it a browser job and get the answer back; the chapter on skills, MCP and tools covers how.

If an agent is not installed yet, Agent Config installs it for you. Settings also shows each agent CLI's installed version against the latest available, so an agent that has drifted behind is visible rather than mysterious.

Bring your own keys

Model Config is a three-layer model: providers (an endpoint plus a credential), profiles (a specific model on a provider, with its reasoning level and token ceiling), and bindings (which profile an agent routes through). Anthropic, OpenAI, Google and any OpenAI-compatible endpoint — OpenRouter included — are first-class.

Model Config, with the providers it ships with
The Mirasim Model Config settings screen, listing the model providers you can connect.

SettingsModel Config lives in Settings, beside Agent Config, Extensions and the traffic monitor. — this control is not on the captured screen.

The buttons in this panel work — click them to look around.

The same three layers over the CLI
# 1 · a provider: endpoint + credential
mirasim ui-cli --port 4970 provider save \
  --kind openai-compatible --name MyRouter \
  --base-url https://example.com/v1 --api-key sk-…

# 2 · a profile: one model on that provider
mirasim ui-cli --port 4970 profile save --provider PROVIDER_ID --model some-model

# 3 · a binding: route an agent through it, or back to its own subscription
mirasim ui-cli --port 4970 model bind --agent codex --profile PROFILE_ID
mirasim ui-cli --port 4970 model bind --agent codex --native

# prove it works before you rely on it
mirasim ui-cli --port 4970 model test --provider PROVIDER_ID --model some-model

The 1M-token window

Some Anthropic models can serve a one-million-token context window. It is never the default — it has to be asked for by choosing the model's long-context variant, marked with a `[1m]` suffix, and Mirasim requests the window through the provider's beta rather than by changing the model itself. Models without a 1M window simply do not offer the variant.

Local models

For work that must not leave the machine, Mirasim can run against local inference: Ollama, plus `mlx-lm` on macOS. Other OpenAI-compatible local servers can be added as ordinary custom providers. Mirasim can probe what is installed, start a backend, pull a model with streaming progress, and delete one when you are done with it.

Local backends from the CLI
mirasim ui-cli --port 4970 local probe                       # what's installed
mirasim ui-cli --port 4970 local start --backend ollama      # bring it up
mirasim ui-cli --port 4970 local pull  --backend ollama --model qwen3
mirasim ui-cli --port 4970 local pulls                       # in-flight pulls

Several accounts, and live quota

You can store more than one account per agent and switch which is active — useful when personal and work subscriptions have different limits, or when a plan resets at an awkward time. The active account for an agent applies agent-wide; sessions do not each hold their own.

Each stored account shows its live quota: the plan, remaining credits, how much of the current windows you have used, and when the limit resets. States are named plainly — healthy, low, exhausted — and you can force a refresh. When the provider's quota API is itself rate-limited, Mirasim says which windows it could still read instead of guessing at the rest.

Agent Config, where accounts and quota live
The Mirasim Agent Config settings screen, showing the signed-in account, its quota, and the update controls.

Cloud quotaHow much of the cloud quota is left, and where turns are being routed. Click to pin it open, hover to preview. — this control is not on the captured screen.

The buttons in this panel work — click them to look around.

Accounts and usage headlessly
mirasim ui-cli --port 4970 accounts --fresh                  # accounts + quota each
mirasim ui-cli --port 4970 usage                              # per-agent quota snapshots
mirasim ui-cli --port 4970 account use --agent claude --id work-account

When quota runs out mid-task

Hitting a limit halfway through a long task is the failure mode that wastes the most time. Sign in to Mirasim's relay and it can pick up the work when your native subscription runs dry — configured by a usage threshold, by the 5-hour or 7-day window, reactively on a refusal, or always on. When a turn goes that way, the transcript labels it: this turn is routing through the quota fallback.

Because the fallback is visible per turn, a session that quietly changed route is not a thing that can happen to you. The terminal surfaces show the same state — the CLI publishes the live route into the agent's own status line rather than drawing over the agent's UI.

Relay state and switches
mirasim ui-cli --port 4970 login              # relay sign-in
mirasim ui-cli --port 4970 relay              # sign-in + fallback state
mirasim ui-cli --port 4970 config set relay.enabled=true relay.threshold=0.8

Something wrong or missing on this page? Tell us