Trusted by 500+ Developers

Stop Juggling
Agent CLIs

Run all your AI agents through one CLI. Access them anywhere via MCP or API. Local execution, cloud control.

Check Our Services See Demo >

System Status

Claude Code

Gemini CLI

Codex

Ollama Models

Mistral Vibe

PUZLD

Unified CLI

Works with your favorite AI providers

OpenAIAnthropicGoogleMistralCohereGroq

Powerful Features

BUILT FOR BUILDERS

Experience the next generation of AI agent orchestration. Local, private, and insanely fast.

Latency

~5ms

Privacy

100% Local

Cost

$0.00/req

Local-First Architecture

Execute agents via local nodes and CLIs. Reduce latency and avoid tight coupling to external APIs.

~/puzld-cli

server.ts

$puzld debate "what is puzldai" -a claude,gemini

@claudeanalyzing query...

@geminiretrieving context...

➜Debate started (session #849)

$puzldai serve

›Starting local agent server...

✓Loaded 5 agents (Ollama, Mistral, Claude Code, Gemini CLI, Codex)

Native Agent Protocol (Puzld MCP)

Built-in protocol that lets tools, CLIs, IDEs, and machines talk to running agents.

IDEs

CLIs

Tools

Puzld MCP Protocol Active

Unified API + CLI

Same agents. Same protocol. Different entry points.

CLI

$ puzld run

id:8f72a

status:running

API

POST /run

id:8f72a

status:running

Multi-Agent Orchestration

Coordinate multiple agents together. Chain outputs, route execution, and let each model do what it’s best at.

5 Agent CLIs

Getting Started

UP AND RUNNING IN MINUTES

01Install the CLI

One command to install the Puzld CLI globally.

$ npm install -g puzldai

02Login & Connect

Authenticate and connect your machine to Puzld Cloud.

$ puzld login && puzld serve --mcp

03Start Building

Your machine is now a Puzld node. Run agents from anywhere.

$ puzld run 'Analyze this code...' -a claude

Feature Deep Dive

AUTOMATION PLAYBOOKS

Build robust workflows tailored to your needs. From simple chains to complex, safe, human-guided operations.

Compare Agent Outputs

Run multiple agents on the same task and compare their outputs side-by-side. Use --pick to have an LLM choose the best response.

Explore Documentation →

~/puzld-compare-mode

# Compare Mode

# TUI (interactive) /compare claude,gemini "Refactor this function" /sequential # toggle sequential execution /pick # toggle LLM picks best # CLI puzldai compare "Refactor this function" \ -a claude,gemini,ollama \ # agents to compare -p # pick best response

Pricing

SIMPLE, TRANSPARENT PRICING

Free

Perfect for trying out Puzld

✓100 requests/month
✓1 connected machine
✓Community support
✓Public agents only

Start for Free

Starter

$7/month

For individual developers

✓3,000 requests/month
✓3 connected machines
✓Priority support
✓Full API access

Upgrade to Starter

Explorer

$39/month

For growing teams

✓20,000 requests/month
✓10 connected machines
✓Dedicated support
✓Audit logs

Scale to Explorer

View full comparison →

What developers are saying

Join thousands of enthusiasts building the future of autonomous agents

View all testimonials →

KnifeFed•14d

Looks cool! Any plans to add support for iFlow and Qwen Code?

SpiritualHorror6594•13d

does it have some main instructions? like claude have claude.md? for various projects? Thanks btw looking great

ElegantAd3211•16d

/debate command? That's a great idea man

Environmental_Help18•16d

This is nuts, I just finished building something similar in n8n but this is far superior

Legitimate-Pumpkin•17d

Looks amazing, man. Thanks for sharing!

DANGERBANANASS•17d

Sounds good bro.

bctopics•17d

Love this!

nsway•17d

I feel like the screenshots were entirely sufficient. I skimmed for 20 seconds and completely understand the value.

InsectActive95•17d

Gem!

Fair-Opposite3871•15d

Looks awesome. Thank you for sharing this!

BrotherBringTheSun•13d

Great idea, right now I literally paste the ideas to fix my code from Gemini into Codex. Will this work without a paid Gemini subscription?

nsway•17d

Does Claude have access to subagents/custom commands/plan mode in all the 'modes' of your project? Thats where I find Opus really shines.

Historical-Lie9697•17d

claude non-interactive mode has a --verbose option which shows text even with pipe output. It also streams json if u give the --output-format stream-json cli flag.

Inex•13d

Thanks mate. I've got subscriptions to Codex, Gemini, and Claude, and this is exactly the kind of tool I was looking for to tie them all together. I'll give it a spin tonight.

InternationalRole724•14d

Nice, will definitly have a look at this, as it is annoying, to switch, as you said.

Scy73•14d

This is very cool and useful. I like the layout and looking forward to how you shape it.

3electron•15d

Does it support DeepSeek API ?

anthonybustamante•16d

Looks cool! I might try this out later. Claude is my favorite, but I have access to codex and Gemini cli so I'd might as well utilize them. This could be a centralized way. I was wondering, does consensus mode burn through tokens too quickly?

C0rtechs•16d

Disregard, works fine under git bash but using cmd (on windows) it doesn't for whatever reason

iichamp•15d

Looks freaking awesome. Would love to see this turned into an MCP. I like running from the claude code cli, but would love to /compare or /debate with a codex model

Stop jumping between agents.
Start orchestrating.

Initialize →View Pricing

Stop JugglingAgent CLIs