Crowe Logic, Inc. v0.9.9

Two flagship models,
one prompt,
one synthesized answer.

Crowe Logic is the agent console for the terminal. One command, a real tool loop, and a model table that names the engine beside every tier. Run two CroweLM tiers side by side on the same question, let a third turn synthesize the answer, and put autonomy levels in front of every tool the agent can touch.

The free tier needs no card and no keys. Pro unlocks the whole CroweLM table through your Crowe ID.

crowe
/dual on
dual mode: on · CroweLM ‖ CroweLM Zenith
What's the best approach for RICO claims on landlord fraud?
CroweLM
RICO requires predicate acts. For your case, you'd need to prove at least two acts of mail fraud, wire fraud, or extortion within...
▊ 312 tok @ 47/s · cached
CroweLM Zenith
Civil RICO (18 USC 1964) is the path. Predicate offenses must form a pattern. The Anti-SLAPP provisions in Arizona don't apply to...
▊ 287 tok @ 52/s
CroweLM Synthesis · merge
Civil RICO under 18 USC 1964 requires a pattern of at least two predicate acts (mail/wire fraud most applicable to landlord cases). Arizona Anti-SLAPP doesn't bar these claims. File in federal court for treble damages...
COST ~$0.087 · CREDITS 15 · TURNS 1 · 2/3 cached
Why Crowe Logic exists

Every other tool of this kind wraps a single model. Crowe Logic runs two concurrently, reconciles them, and delivers one answer you can trust more than either alone.

Features

Built for operators who need more than chat.

Dual-model concurrent

Run CroweLM and CroweLM Zenith side-by-side on the same prompt. Watch both stream in real time. Compare reasoning paths before committing to an answer.

Synthesis layer

A third model reads both responses and produces the final synthesis. Three modes: merge (blend strongest claims), judge (pick winner), diff (show gaps).

Autonomy tiers

Read only, edit, execute, full. The level filters which tools the agent can call at all, and it is process-wide, so the execution gate and the tool list can never disagree.

Live cost visibility

Every turn shows tokens, cache hits, USD cost, credits consumed. You always know what a conversation is actually burning.

/replay and /fork

Re-run any past turn on a different model. Fork a conversation to try a different approach without losing context. A/B models like flipping a switch.

Five families, three rungs

Every family answers on the same edge chain, each at normal, high or max effort. An Auto router picks the family for the task, and a paywalled pane falls back to one the plan can reach.

The CroweLM family

Five families. Pick by the work.

One name per job, each at three effort rungs, plus an Auto router that picks for you and a free Mycelium tier for cultivation work. The name stays put when the model behind it moves, so your pins and scripts do not have to.

Crowe Logic does not train these engines. Each family runs on a named frontier model hosted on Microsoft Azure, with our grounding, knowledge base, tools and orchestration on top. Ask any tier what it runs on and it will tell you.

CroweLM
The default
Where a turn starts. Your knowledge base and domain data are mounted into the session, and it reads a screenshot as readily as a paragraph.
Runs on Claude Fable 5 by Anthropic
Zenith
Decisive analysis
For the question the default answered vaguely. Sharp, structured, willing to commit. The other half of the dual-mode pair.
Runs on Claude Opus 5 by Anthropic
Depth
Long horizon
Multi-step work over very large contexts, with twice the thinking room at every rung.
Runs on Grok 4.6 by xAI
Coder
Code in your codebase
Follows the patterns already in your project instead of importing its own.
Runs on GPT-5.6 by OpenAI
Flash
Drafts and feedback loops
Brief and action oriented, for tight iteration. Lowest credit cost per turn.
Runs on DeepSeek V4 Flash
CroweLM Auto
Routes per turn
Reads the task and picks the family, so a bare crowe already lands on the right tier. Vision, Grower and Kernel personas sit beside it, each on Claude Fable 5.
Frontier picks, by name
Your own inference keys
GPT-5.5, GPT-5.4, DeepSeek V4 Pro, Kimi K2.7 Code, Kimi K3, GLM 5.3 Flash and Claude Opus 4.8 sit in the same /model table under their own names. They carry no CroweLM identity and introduce themselves as what they are.
CroweLM Mycelium
Free tier. The one Crowe trained
A Gemma 4 fine-tune on Crowe Logic's cultivation corpus, serving the free tier at twenty turns a day. This is the model, the dataset and the knowledge base the CroweLM name was built for.
Pricing

Start free. Pay for the table.

The free tier is real and needs nothing from you. Pro is one price for every CroweLM tier, every effort rung, and the desktop and web apps signed in with the same Crowe ID. Credits meter usage so a vendor rate-card change never reaches your invoice: flagship turns cost 5, balanced 2, fast 1, and dual mode sums both sides.

Free
No card, no keys
$0
  • CroweLM Mycelium, 20 turns a day
  • The full tool loop and autonomy levels
  • Frontier picks on your own keys
  • Daily self-update
Install
Team
3 seats and up
Per seat
  • Credits pooled across the team
  • Shared workspaces
  • Admin cost reports
  • Everything in Pro, per seat
Talk to us
Enterprise
25 seats and up
Custom
  • Dedicated endpoints
  • Data retention controls
  • Audit logs
  • Self-hosted node option
Talk to Crowe Logic
Already hold inference keys?
The frontier picks run on your own keys inside the same console, on every plan including free. Point crowe init --node at your endpoints and the picks appear in /model under their own names. No separate tier, no credit meter on those turns.
Install
Install

Two commands and you're running.

pip install crowe-logic
crowe

Python 3.10 or newer, on macOS, Linux or Windows. First launch runs on the free tier with no card and no keys: CroweLM Mycelium, twenty turns a day. crowe login signs in with Crowe ID and opens the rest of the table.

Or the app

The same operator, with a window.

Crowe Logic desktop runs the same agent against the same gateway, and adds the things a terminal cannot draw: reviewable edit diffs you approve or reject, a real shell, a file tree, git, and an in-app browser. Signed and notarized on macOS. It updates itself.

Apple Silicon and Intel disk images, AppImage and .deb. Windows is next.

FAQ

Honest answers to the obvious questions.

Why would I pay $99 when Claude Pro is $20?

Claude Pro gives you one model in one interface. Crowe Logic Pro gives you five CroweLM tiers running on Claude Fable 5, Claude Opus 5, Grok 4.6, GPT-5.6 and DeepSeek V4 Flash, in one console, with a real shell, a registry of more than a hundred tools spanning web search, browser automation, MCP and public records, and dual-mode synthesis across two of them at once. If you only need chat, Claude Pro is fine. If you are doing research, writing code or running investigations, the table earns its price. Start on the free tier and decide from inside the console.

What are credits and how do they work?

Credits decouple what you pay us from what we pay upstream providers. A fast-model turn costs 1 credit, balanced 2, flagship 5. Dual mode sums both sides. Synthesis adds 5. Pro's 3,000 credits support roughly 600 flagship turns or 300 dual-mode turns a month. You can see exact cost and credit consumption live in the HUD as you work.

What models actually run underneath CroweLM?

Named frontier models, hosted on Microsoft Azure. CroweLM runs on Claude Fable 5 and Zenith on Claude Opus 5 (both Anthropic), Depth on Grok 4.6 (xAI), Coder on GPT-5.6 (OpenAI), and Flash on DeepSeek V4 Flash. Crowe Logic does not train these engines; what we build is the routing, the grounding in your knowledge base, the tool registry, the dual-mode synthesis and the operator workflow around them. The family names stay fixed so your pins survive when we move a family to a stronger engine, and any tier will tell you what it runs on if you ask.

Can I use my own API keys?

Yes, on every plan. The frontier picks (GPT-5.5, GPT-5.4, DeepSeek V4 Pro, Kimi K2.7 Code, Kimi K3, GLM 5.3 Flash, Claude Opus 4.8) run on keys you hold, and crowe init --node points the console at endpoints you host. Those turns are billed by your provider, not metered by us. The CroweLM tiers run through your Crowe ID.

What stops the agent running something it shouldn't?

When an agent needs to run shell commands (install dependencies, test code, probe a remote server) there is an implicit trust decision. Tools run on your machine, so the decision is made before the call, not after: the autonomy level - read only, edit, execute, full - filters which tools exist for that turn. A read-only session cannot call a writing tool, because the tool is not in the list it was given. The level is process-wide, so the execution gate and the tool list can never drift apart.

Is my data trained on or sold?

No. Crowe Logic doesn't send your conversations to any training pipeline by default. A turn goes from your CLI to the edge that serves the family you picked and back. That edge has its own data policy, which you can read on its site. On Enterprise plans you get additional data retention controls and optional on-premises install.

Can I cancel anytime?

Yes. Plans are billed monthly and cancel at the end of your billing period. No lock-in, no cancellation friction. Your conversation history remains in your local install regardless.

Who builds this?

Crowe Logic, Inc., founded by Michael Crowe. We ship the CLI as our primary product. The same framework powers other Crowe Logic agents (cultivation advisor, research, compliance) sold to adjacent markets. We're a bootstrapped company, not VC-backed. The product exists because we needed it for our own work.

Start today

One command. Every engine named.

Install, type crowe, and you are on the free tier with no card and no keys. Pro is $99 a month for the whole CroweLM table. Monthly billing, cancel any time, your history stays on your machine.

Checkout runs through Crowe ID and Stripe and provisions immediately.