{{ ambient }}
Plexar
{{ host }}
From multiplexer to platform

Give yourself back time.

Plexar started as a terminal multiplexer. It's now a platform: our models, our agent harness, and the tools to build with them or just use them.

Get Studio open source
plexar-studio — 4 panes
● {{ p.title }}
{{ l }}

One platform, nine surfaces.

Studio and Framework are open source, Mobile is a free Android download, and Harness is a Windows download. Using the rest, and Harness itself, is by invite for now; each tool is granted on its own.
{{ p.badge }}
{{ p.name }}
{{ p.line }}
{{ p.sub }}.plexar.tech
{{ s.n }}
{{ s.t }}
{{ s.d }}

See each tool before you ask for it.

A look at every Plexar surface as it is today. Studio and Framework are open source; Harness and Mobile are downloads. Access to the rest is by invite.

Plexar {{ sel.name }}
{{ sel.sub }}.plexar.tech

{{ mTitle }}

Signed in as {{ userName }}
Endpoint
Stays the same when the model behind it changes.
https://llm.plexar.tech/v1
Your key
Admits exactly what you were granted, nothing wider.
{{ keyStatus }}
{{ keyShown }}
Shown once. Store it somewhere safe.
{{ k.id }}{{ k.when }}
Connect
OpenAI-compatible. Wherever a client asks for an API key, paste your key; wherever it asks for a base URL, paste the endpoint.
{{ code }}
Authorization: Bearer YOUR_KEYis sent with every request.
Playground
+
Qwen3.6 35B ⌄
Temp 0.7 ⌄
↑
{{ s.l }}
{{ s.v }}
Tokens, last 14 days
OutputInput
Sep 11Today
Plexar Framework

Agents propose. People approve. The gate decides.

The control plane for agent runs. Prompts go into a bucket, a batch is picked with a stated reason, a named person approves it, and a daemon runs it. A task lands only if the gate exits 0.

$ plexar init && plexar up
One task, start to finish
bucket: web · task t-142
{{ s.n }}
{{ s.t }}
{{ s.d }}
{{ s.n }}
{{ s.t }}
{{ s.d }}
~/plexar-agents/logs/2026-09-24.jsonl
{{ r.line }}

A review chain that shows its evidence.

A worker does the task, a verifier runs it and photographs anything visual, and an approver checks the verifier. You get a QA report with the screenshots and every model's verdict.

It is still AI judgement. Treat each verdict as a lead to confirm; your own verdicts are kept beside the AI's and make the next report sharper.

{{ c.n }}
{{ c.t }}
{{ c.d }}
{{ c.v }}
{{ r.t }}
{{ r.d }}
{{ r.t }}
{{ r.d }}
Plexar Chat
Search chats, projects, messages⌘K
D
Signed in
Where should we start?
Ask a question, paste something to work through, or pick a starting point.
{{ c.t }}
{{ c.d }}
Ask Plexar anything
+
Attach
bonsai-27b ⌄
↑
Plexar can be wrong. If an answer stops early, the message will say so.
9:41●●● 5G
PLEXAR
Good afternoon
Assistant
Plexar Assistant
Idle
Your morning summary is ready.
Open chat
Plexar Studio
Signed out. Tap to sign in.
›
Plexar Chat
›
+ Add a tool
Plexar Access

Models and every Plexar tool are granted by hand. Pick what you need and tell us what you're building.

{{ reqCount }} requested
Reviewed by hand
Request received.
You'll get an email once it's reviewed. Approved tools switch on at their own subdomains, and your usage shows up here.
{{ s.n }}
{{ s.t }}
{{ s.d }}