Building on Blue Mesh starts with a deployment, not a signup
There is no self-serve surface here: nothing to sign up for, and no public developer portal behind this page. The full list of what does not exist is the first band below, unhedged and undated, because a reader who came looking for documentation should hit that before anything else asks for their attention. What does exist is a different arrangement from an API key, and the rest of the page is the mechanism of it: where the platform runs, what your team builds on, what a workflow reaches, and what you can see while it runs.
No public developer portal, and no date for one
Stated once, at the top, so it never has to be inferred from silence further down. None of the following exists today and nothing on this site links to any of it. There is no date attached because none has been set, and a quarter nobody has committed to would be worth less to you than this paragraph. Absent here means not published: it does not mean the platform cannot do the thing, and the four bands below are what it does.
- No documentation site, no getting-started guide, no tutorials
- No API reference, no OpenAPI artifact, no SDK, no sample project
- No sandbox, no trial tenant, no self-serve signup
- No Discord, no forum, no public repositories, no showcase program
- No help center, no knowledge base, no ticket portal, no status page
The developer experience is an engagement, and it has a shape
Four stages, and not one of them is a form. This is what a team building on Blue Mesh does, roughly in the order they do it, and it is the reason the absences above are absences rather than oversights.
It deploys into your environment
Three modes, and they are private cloud, on-premises, air-gapped. The platform goes where your data already lives, so your network controls, identity provider and change process apply unchanged. There is no public instance, so no API key to issue.
Workflows are built on a canvas
BM Studio is a drag-and-drop canvas, and the brochure's claim is that it needs no machine learning hires. A Leader Agent decomposes each task and directs specialist sub-agents end to end. Workflows nest, so one that already works becomes a step inside a larger one.
Or you describe it and it gets built
BM Architect takes a workflow described in plain language and designs, tests and deploys it. What comes back lands on the same canvas as step 02, so a described workflow and a drawn one are the same object. The closest thing to a quickstart, and it runs in a conversation.
You work with the people who wrote it
The team deploying Blue Mesh works with its engineers directly, on a channel agreed for that deployment. The public route is the contact form, read by a person. No ticket number, no queue, because there is no ticketing system behind it. This page is not your support channel.
The extension surface, named
The question a developer arrives with is what a workflow can reach and what happens when it cannot reach something. Four answers, and each one is a mechanism rather than a claim about openness.
Integrations
Agents act through the 300+ systems and tools you already run, under the same credential discipline your team already enforces. Nothing gets rebuilt in our image to make a workflow possible, which is what keeps the first deployment a configuration exercise rather than a migration.
Custom tools and MCP
A tool your agents do not have is a tool you add: external APIs plug in as reusable tools, and BM Studio speaks Model Context Protocol, so an MCP server your team already runs is reachable from a workflow without waiting on a release from us. The catalog is a starting point, not the ceiling.
Inbound webhooks
Incoming webhook endpoints receive and process events as they happen, which is how a system you already run starts a workflow instead of a workflow polling for it. Inbound is the direction that exists today, and this page will not imply a second one.
OpenAI-API compatibility
A Studio deployment can be addressed as an OpenAI-compatible API, so code already written against that shape runs against your own deployment without modification, and OpenRouter connects the same way when you want several models behind one interface. Migrating in costs less when the client library does not change.
What you can see while it runs
Debugging an agent is harder than debugging a request, because the interesting failure is a decision rather than a stack trace. These three surfaces are what make such a decision reviewable after the fact.
The step-by-step trace
Agent actions render as they happen: the reasoning, the tool called, the status of each step. When a workflow surprises you, you read which step chose it.
Token tracking
Token use is tracked per request, so cost is a readout your team watches while tuning. No figure is published here, because none of yours is measured yet.
The audit record
Every decision and result is logged for compliance audit inside your own environment, under your retention rules. It is the record your security reviewer asks about.
Who this does not suit today
If you want to read the reference tonight, install it on your own laptop this week and reach a decision without talking to a human being, Blue Mesh is not that product yet, and calling the contact form self-serve would not make it one. Every deployment currently goes through a conversation, which is slower than an API key and is a fair description of what you are choosing. When the documentation exists, this page is where it will appear, and it will appear without a countdown in front of it.
- Changelog The one developer artifact that is published: releases, feature changes and fixes, one entry per product.
- BM Studio The canvas from step 02 in full, including the Studio and deployment feature lists this page summarizes.
- BM Architect Step 03 as its own page: a workflow described in plain language, then designed, tested and deployed.
- All features The capability hub, and where the connectivity and observability surfaces above are each written out.
- Trust & Security What your security reviewer will ask for: the three deployment modes, the named controls, and what we do not claim.
Bring the workflow you would have built against an API
Describe what you would have wired up yourself. We will show you where it lands on the canvas, what it connects to in your own stack, and which parts your team still owns after it deploys.