AI Workflow™ system

Ari™ turns requests into finished work.

Meridian™ is the workflow system behind YAKKL® Ari™. It classifies the request, builds a plan when needed, routes work to the right model and AI vendor, saves the evidence, creates artifacts, and keeps the whole run searchable after the session ends.

Meridian™ Founding Builder Preview

Early signup starts NOW. Downloads open June 2026. Founding pricing ends September 30, 2026.

Early builders get influence, founding economics, and access to a product designed to finish real work instead of trapping it inside a transcript.

Fast path

Early signup starts NOW. Downloads open June 2026.

Sign up now to lock founding pricing. CLI and Desktop downloads open in June. Move into the right plan when the value is clear.

Turn any AI request into plans, artifacts, audits, and finished work
Ari is the guided experience; Meridian is the execution system underneath it
Route phases across frontier, fast, local, and visual model types with multi-vendor support
Founding Builder Preview is open while the product hardens in public

AI Orchestration

Meridian™ is built to define AI Orchestration as its own market.

The winning layer is not one model, one chatbot, or one coding surface. It is the system that knows which model to use, what state to preserve, what evidence to keep, and how to govern the work.

Model control

Route work by task, not by vendor habit

AI Orchestration treats every model as a specialized engine. Meridian selects for reasoning depth, speed, coding strength, visual work, cost, privacy, and output quality.

Workflow state

Keep the plan, artifacts, audits, and handoffs together

The market is bigger than chat. Serious AI work needs state, phases, evidence, replay, budget controls, and a durable record that survives model changes.

Governance

Make AI work inspectable before it becomes operational risk

Meridian gives buyers a control plane for model routing, CU policy, BYOK fallback, review gates, and proof of what happened during execution.

Three Steps

Plan the work. Route the models. Keep the evidence.

Meridian sells best when the path stays simple: prove the product, pick the right plan, and move forward without buying more governance than needed.

Step 1

For work that cannot die in a transcript

Meridian turns raw intent into structured delivery with plans, phases, handoffs, artifacts, memory, audits, and replay so the result stays usable after the first run.

Step 2

Choose the plan that matches how you build

Free and Builder tiers serve individual builders. Teams adds shared execution. Business adds cleaner control and annual purchasing. Enterprise handles governed scale and custom routing.

Step 3

Lock founding economics in for a limited time while the early window is open

Join from NOW through September 30, 2026 to lock in founding pricing, preview influence, and eligible CU bonuses while the program remains open.

01

Start with an AI request and optionally convert it into a plan when the work seems too large for chat

Meridian can turn any AI request (even /skill level requests) into a structured plan (if needed) with phases, sub-phases, artifacts, dependencies, audits, and handoffs before expensive work starts.

You can start with an AI request (even /skill level requests), Meridian will let you know if a full extensive plan is the best fit or not. If not, you can just run the request and get the results. Waste less later.
02

Route each phase or request to almost any vendor model that fits the job

Use deeper reasoning where it changes the result, faster models where speed matters, and local or visual models where privacy or media generation matters more than generic chat.

We suspect "There is NO WHERE" that provides scored system prompts for different requests or workflows. The requests can determine which model types are the best fit for the job and which vendors to use (Anthropic, OpenAI, Google, etc.). Different vendors have different strengths and weaknesses and Meridian is the first to provide a way to route the work to the best fit for the job. Many specialized model types for almost any request or workflow.
03

Keep the work visible after the first run

Plans, handoffs, artifacts, memory, and replay stay visible so teams can inspect what happened, reuse what worked, and control what changes next.

The work survives the session.

Plans

Choose the plan that fits the way you build.

Start with the plan that matches your current workflow, then move up only when shared execution, pooled CU, stronger model routing, or stronger control becomes a real operational need.

Individual

For builders who need more than simple chat and want to use multiple models and vendors without doing a full team rollout

Free gets you started, with limited features, if you only do a few requests a month. Builder adds stronger workflow implementation. Pro+ is the best individual option - you can use all the models and all AI vendors with multi-model execution plans.

Team

For teams that need shared execution instead of isolated sessions

Teams adds pooled CU, shared execution, dashboards, and visibility across the group so work stops disappearing into private prompts and disconnected sessions.

Business

For companies that need cleaner control without enterprise drag

Business adds annual purchasing, stronger auth, SSO, RBAC, and security controls for companies ready for disciplined rollout but not full enterprise contracting.

Governed

For governed scale and custom operating rules

Enterprise is the route for policy, compliance, dedicated routing, org hierarchy, and custom commercial structure once Meridian becomes governed infrastructure.

Control proof

Three things serious builders want to know before they commit.

Which model did the work, why was it selected, and what evidence was produced. Meridian keeps those answers visible.

Model & AI Vendor Fit

Model choice and AI vendor fit are explicitly routed

Meridian routes work by capability: deep reasoning, coding, visuals, extraction, review, local privacy, cost, latency, and quality class. Different vendors have different strengths and weaknesses and Meridian is the first to provide a way to route the work to the best fit for the job. Many specialized model types for almost any request or workflow.

Inspectability

See how the work was done

Rhumb™-aligned Master Plans, handoffs, audits, replay paths, and artifacts stay visible so teams can inspect how Meridian got to the result instead of trusting a black box.

Preview Honesty

Early access is clear about product maturity

Founding Builder Preview says the product is still hardening. Early users get influence and better economics because they are helping shape the operating surface. So, signup now and get in on the ground floor. Also, subscribe to our newsletter to get the latest updates and news about Meridian.

Visual proof route

See how a prompt becomes governed delivery.

See classification, model routing, artifacts, audits, handoffs, and searchable recall before you choose the plan that fits your team.

Open orchestration flow

Open protocol

Rhumb Workflow Protocol™ is the standard. Meridian™ is the productized implementation.

Rhumb Workflow Protocol™ defines the portable workflow contract: intake, plans, manifests, state, handoffs, schemas, lifecycle rules, and conformance. Meridian™ uses that open standard to orchestrate models, produce artifacts, audit work, and keep delivery inspectable.

Protocol

Plain-file workflow that can outlive any one tool

Rhumb Workflow Protocol™ is Apache-2.0 and protocol-first. It gives teams a shared language for AI-assisted work without forcing them into a private project format.

Reference

Meridian proves the protocol in real execution

Meridian™ turns Rhumb Workflow Protocol™ concepts into managed orchestration: Master Plans, model routing, artifacts, audits, handoffs, replay, and governed delivery paths.

Ecosystem

Open standard first, commercial acceleration second

Builders can study Rhumb Workflow Protocol™ at rhumbprotocol.dev. Teams that want the operating system around it can use Meridian™ instead of assembling every workflow surface themselves.

Rhumb Workflow Protocol™: rhumbprotocol.dev

CU blocks

Credit Units buy managed execution, not raw tokens.

A Credit Unit, or CU, buys a unit of managed work rather than a raw token bucket. Meridian-managed execution lets the system optimize compression, handoffs, cache-aware routing, and execution strategy so teams can get more finished work from every refill.

Managed execution starts with guided preview outcomes in Free, expands in Builder, opens fully in Pro+, and scales into pooled inventory for Teams, Business, and governed volume for Enterprise.

Scroll right to inspect the full individual CU range and where Builder stops.

Individual CU access

Free starts with limited features and can buy Starter and Small CU. Builder adds Medium. Pro+ unlocks the full individual CU range for heavier preview work.

Guided preview

Starter Block

The fastest way to get to something tangible. Ideal for landing pages, small app shells, campaign assets, first-pass content, and quick proof of concept work. CUs are units of work, not raw token buckets.

Price $25
Credit Units (CU) 500 CUs (Credit Units)
Expiry 60-day expiry

BYOK and managed keys

BYOK and managed execution are two different operating models.

Choose direct provider control, optimized managed execution, or a hybrid path that falls back cleanly when CU runs low.

BYOK

Bring your own keys (BYOK) when you want direct provider billing

BYOK lets Meridian run through your own model accounts so you keep direct provider billing and account-level visibility.

  • Change to BYOK at any time.
  • Best for buyers who want direct vendor billing and provider-level visibility
  • Available in every tier, including Free
  • Useful as fallback when managed CU inventory is low - you can use both at the same time
Managed

Use Meridian-managed execution when you want a easier path

CUs use Meridian-managed keys so Meridian can optimize the execution path for better workflow density, cleaner handoffs, and less operational friction.

  • Credit Units (CUs) are units of work, not raw token buckets
  • Built to produce more usable work per spend than simple pass-through usage
  • Ideal when you want less setup friction and faster starts - default setting
Hybrid

Blend managed CUs and BYOK when you want flexibility

Hybrid mode lets Meridian use CUs first while keeping your own keys available when you want direct-provider access or a clean fallback path. This means if run out of CUs, you can still run the request with your own keys with no downtime.

  • CUs-first keeps the easiest path in front of the you or your team
  • You or your team can still prefer your own keys when you want BYOK priority
  • Useful for teams that want resilience without forcing one billing model
CU-first Meridian-managed execution
Plan enters Meridian Processing Pipeline
Compressed plan

Optimized handoffs, compressed context, cache-aware routing, model processing optimization, and a denser unit-of-work output path.

Typical result Lower effective token load
Pass-through BYOK (Bring Your Own Keys)
Plan enters Meridian™ Processing Pipeline
Standard plan

Full prompts, fuller handoffs, direct provider billing, and less opportunity for system-managed optimization.

Typical result Standard provider token load

Managed execution is designed to produce denser work per dollar through compression, routing, and cache-aware execution. BYOK remains available whenever direct provider control matters more. CU costs may change as vendor pricing, model availability, capacity, cache economics, and abuse controls change.

Decision point

Once the workflow is clear, the buying decision gets simple.

Individuals choose Free, Builder, or Pro+. Teams add seats and shared execution. Business adds annual purchasing, stronger auth, and security controls. Enterprise moves into a direct evaluation path for governed deployment.

If two or three people mostly work independently, multiple Pro+ seats can still be the better value. Move into Teams when shared execution, pooled CU, shared sessions, or manager visibility become the clear requirement.

Compare plans

License confirmation

Free, Builder, Pro+, Teams, Business, Enterprise.

Choose the plan that matches the way you want to work now, then move up only when you need more throughput, coordination, or governance.

Free starts with guided preview, open-source workflow, BYOK, and limited CU access. Builder and Pro+ offer monthly and annual pricing, with Pro+ as the strongest individual plan. Teams starts at two seats. Business starts at five seats and is annual only. Enterprise stays custom. CU purchases happen inside signed-in Meridian surfaces.

Community

Free

$0

If you only need limited features. Workflow implementation and the option to keep running with BYOK or small CUs purchases or both. It does have a few limitations, but it is a good way to get started if only do a few requests a month.

  • Open-source Rhumb Workflow Protocol™ (RWP) based workflow implementation
  • BYOK remains available for fallback
  • Starter and Small CU access
  • Basic plans and local workflow - CLI only
  • Single model chat - Anthropic, OpenAI, Google, etc. - pick your favorite model vendor - one vendor for all requests. You still have access to all the models from the vendor you choose.
  • Session persistence - you can have a session running for up to 2 days - this is a limitation of the free plan
  • Up to 2 active plans - you can have only 1 active plan at a time executing at a time - this is a limitation of the free plan
  • Only maintains history for the last 14 days - this is a limitation of the free plan
  • No desktop access - this is a limitation of the free plan
  • No optimizations nor request shaping, routing or scoring - this is a limitation of the free plan
  • Free to use forever - no credit card required unless you want to purchase CUs

Choose your path

Pick the next step and move.

Start with a guided preview, open self-serve checkout, or move into enterprise evaluation without retracing the whole page.