Skip to content

AIagentsthathandleit.

You just decide.

Chain agents together, run them on your machines, and pay a flat rate. They do the work. You steer.

mentiko — New Decision
1prompt
2preferences
3options
4tasks
5chain
6agents
What do you want to build?
Add notification preferences storage
Your own instance
Bring your own keys
Event-driven
JSON-defined

Threestepstoautomation

01
Define your chain
Drag agents into a chain or write the JSON yourself. Connect them with events. That's the whole setup.
02
Bring your own keys
Claude, GPT-5.4, Gemini, local models -- whatever you use. Your keys stay on your machine. We never see them.
03
Run on your schedule
One-off, cron, or webhook-triggered. Chains handle retries on their own and alert you only when a decision is needed.

Buildachain.Runitonyourschedule.

End-to-end automation

Define what needs to happen. Mentiko handles the rest. Agents run in sequence or parallel, retry on failure, and alert you only when a decision is needed.

Your infrastructure

Your own instance, not a shared tenant. API keys never leave your machine. All state is JSON you can grep and git-commit. Just export and leave anytime.

Up and running in minutes

Visual builder or raw JSON. Real-time monitoring, watchdog for stalled runs, retry with backoff. Build a chain today and it works tonight.

AI-guided
decision making.

Not every task has a clear answer. The decision flow presents tradeoffs as simple A/B choices, learns your preferences, and generates tailored options with an execution plan.

1Swipe through preference tradeoffs
2AI generates tailored options
3Pick one and get an execution plan
mentiko — Decision Flow
1preferences
2options
3plan
Preference 1 of 7
Optimize for development speed
or long-term maintainability?
architecture
Ship fast now
rapid_iteration
A
Build for scale
long_term_stability
B
Skip (S)
arrow keys: left = A, right = B, up = back

Whythisexists

Most platforms put you in the labor seat -- writing glue code, managing retries, babysitting agents. We think that's backwards.

AI is good enough to orchestrate itself. Your job is to decide what matters, not to do the work. That belief shapes everything we build.

We'd rather move slow than break your trust.
We'd rather impress you with simplicity than complexity.
We'd rather you choose to stay than have to.
We'd rather you steer than pedal.
0+
agents available
0.0s
avg chain start
$0
flat-rate / month
0
API keys exposed

HowMentikocompares

MentikoOthers
InfrastructureYour own instanceShared cloud
API keysNever leave your infraPassed through theirs
PricingFlat-rate, unlimitedPer-execution / per-seat
PortabilityJSON + GitProprietary format
ArchitectureEvent-drivenLinear steps
1,000 runs/mo$29$500+

$29/month.Unlimitedeverything.

Your own instance. Unlimited chains, runs, agents, schedules. No per-execution fees. No per-seat pricing.

$29/mo
2 GB RAM / 1 vCPU
Per-execution platforms cost $500+ for 1,000 runs. Same workload here: $29.

faq

Questions we actually get asked

An event-driven platform for orchestrating AI agent pipelines. You define chains of agents that research, write, review, analyze -- whatever you need automated. Agents trigger other agents through file-based events. Think CI/CD, but for AI workflows.

Those are Python libraries. Mentiko is a full platform -- visual chain builder, real-time monitoring, scheduling, multi-tenancy, RBAC. Agents run in real terminal sessions, not sandboxed code execution. And you own the infrastructure. No vendor lock-in.

Any model you have API keys for. Claude, GPT-5.4, Gemini, Mistral, local models via Ollama -- Mentiko doesn't care. Bring your own keys. They never leave your infrastructure.

Your data stays on your instance. API keys are encrypted at rest with AES-256-GCM. Nothing transits our servers. You can grep your own state files. Full data export anytime.

Yes. Mentiko runs on any VPS or bare metal. Docker support included. You get an isolated instance, not a shared tenant. Your machine, your rules.

Flat-rate monthly. No per-execution fees, no per-seat pricing, no usage meters. You pay for compute resources, not per API call. Run 10 chains or 10,000 -- same price.

We're onboarding in batches. When it's your turn, you'll get an invite code, spin up your instance, and start building chains.

Not necessarily. The visual chain builder lets you drag-and-drop agents and connect them. But if you prefer, everything is JSON -- you can write chains in your editor and git-commit them.

We'reonboardinginbatches.

Every account gets a real machine, not a shared quota. Leave your email and we'll let you know when it's your turn.