Manifest Studio

---

Manifest Studio is the heart of ContextMesh: a fully visual, drag-and-drop environment for building MCP manifests without hand-coding JSON. It combines:

• Canvas – your grid-based workspace

• Block Palette – six block/node types

• Properties Panel – edit each node’s settings

• Live JSON Preview – real-time manifest output

• Validation Terminal – one-click manifest checks

---

1. UI Layout

flowchart LR
  subgraph LeftSidebar [Node Palette]
    Palette[Block Palette]
  end
  subgraph CanvasArea [Canvas]
    Canvas[Visual Canvas]
    Zoom[+ –  ☐ 🔒]
  end
  subgraph BottomArea [Validation Terminal]
    Terminal[▶ Validation Log (collapsed)]
  end
  subgraph RightSidebar [Properties & JSON]
    Run[► Run Validation]
    Props[“Select a node…” Properties Panel]
    Preview[Live JSON Preview]
  end

  Palette --> Canvas
  Canvas --> BottomArea
  Canvas --> RightSidebar

• Left Sidebar – drag any of six node types onto the canvas

• Center Canvas – pan, zoom, snap-to-grid, arrange your manifest graph

• Bottom Terminal – expand for timestamped validation logs

• Right Sidebar – node properties editor, live JSON, and “Run Validation”

---

2. Canvas Overview

1. Drag blocks from the palette onto the grid.

2. Move and snap them into a tidy graph.

3. Connect with edges to define data flow.

4. Zoom or pan to navigate complex manifests.

---

3. Block Palette

Six core node types—color-coded by category:

flowchart LR
  subgraph Palette
    Meta([Manifest Metadata]):::meta
    Cap([Capabilities]):::cap
    Res([Resources Container]):::node
    Tool([Tools Container]):::node
    Prompt([Prompts Container]):::node
    Samp([Sampling Container]):::node
    Roots([Roots Container]):::node
    Trans([Transport]):::node
    Conn([Connector Edge]):::edge
  end

  classDef meta  fill:#8AE8FF,stroke:#005f8a;
  classDef cap   fill:#A3E8FF,stroke:#006f8a;
  classDef node  fill:#CFA3FF,stroke:#5c005f;
  classDef edge  fill:#FFCF8A,stroke:#8a5f00;

• Manifest Metadata

• Capabilities (toggle which containers appear)

• Resources (file-like data definitions)

• Tools (functions the model can call)

• Prompts (template messages)

• Sampling (LLM request settings)

• Roots (workspace roots/URIs)

• Transport (stdio, SSE, POST)

• Connector Edge (directs data flow between nodes)

---

4. Node Types & Properties

Each node’s editable fields appear in the Properties Panel when selected.

4.1 Manifest Metadata

Property	Type	Required	Description
name	string	✔	Unique server identifier
version	string	✔	SemVer (e.g. “1.0.0”)
description	string	✘	Brief human-readable summary
author	string	✘	Author or organization name

4.2 Capabilities

Property	Type	Required	Description
resources	toggle	✘	Enable the Resources container
tools	toggle	✘	Enable the Tools container
prompts	toggle	✘	Enable the Prompts container
sampling	toggle	✘	Enable the Sampling container
roots	toggle	✘	Enable the Roots container

• Enabling a capability spawns its container node on the canvas.

4.3 Transport

Property	Type	Required	Description
type	dropdown	✔	stdio
sseEndpoint	string	✘ (SSE)	Path for SSE GET (e.g. /sse)
postPath	string	✘ (POST)	POST endpoint (e.g. /messages)

---

5. Containers

When a capability is enabled, drag its container onto the canvas. Below are each container’s mandatory “properties” fields and how they serialize in JSON:

5.1 Resources Container

• properties must include:

  • resources: array of

    • uri (string)

    • name (string)

"resources": [
  {
    "uri":  "file:///path/to/file.txt",
    "name": "File Description"
  }
]

5.2 Tools Container

• properties must include:

  • tools: array of

    • name (string)

    • description (string)

    • inputSchema:

      • type: "object"

      • properties: { … }

"tools": [
  {
    "name": "calculate_sum",
    "description": "Add two numbers",
    "inputSchema": {
      "type": "object",
      "properties": {
        "a": { "type": "number" },
        "b": { "type": "number" }
      }
    }
  }
]

5.3 Prompts Container

• properties must include:

  • prompts: array of

    • name (string)

    • description (string)

    • template (string)

"prompts": [
  {
    "name":        "GreetUser",
    "description": "Say hello",
    "template":    "Hello, {{username}}!"
  }
]

5.4 Sampling Container

• properties must include:

  • sampling: object

    • modelPreferences (optional)

    • temperature (number)

    • maxTokens (integer)

"sampling": {
  "temperature": 0.7,
  "maxTokens":   200
}

5.5 Roots Container

• properties must include:

  • roots: array of

    • uri (string)

    • name (string)

"roots": [
  {
    "uri":  "file:///home/user/project",
    "name": "Project Root"
  }
]

---

6. Connector Edge

Property	Type	Required	Description
from	node ID	✔	Source node’s output port
to	node ID	✔	Target node’s input port

• Edges drive the JSON generator to nest definitions properly.

---

7. Properties Panel & Inline Validation

• Select a node to reveal its form fields.

• Required fields are marked with “*”.

• Invalid inputs highlight red and display a tooltip.

• JSON-schema fields (e.g. inputSchema) validate on the fly.

---

8. Live JSON Preview

As you build, the right-sidebar preview streams out a fully compliant MCP manifest:

{
  "serverInfo":   { … },
  "capabilities": { … },
  "transport":    { … },
  "resources":    [ … ],
  "tools":        [ … ],
  "prompts":      [ … ],
  "sampling":     { … },
  "roots":        [ … ]
}

• Click Copy or Download to export manifest.json.

• Highlighting follows your canvas selection for easy debugging.

---

9. Run Validation & Log

• Run Validation executes a suite of rule-based checks:

  1. JSON syntax

  2. Metadata presence

  3. Unique IDs & names

  4. Each container’s required fields

  5. Edge rules & port matching

  6. No extraneous fields

• Entries appear timestamped; failures in red.

• Final tally: “X passed, Y failed.”

• Collapse with the “–” button.

---

10. Quick-Start Example

1. Add Manifest Metadata → set name, version.

2. Enable Capabilities → drag Prompts + Sampling.

3. Define a Prompt → enter name, template.

4. Configure Sampling → choose model, temperature, maxTokens.

5. Wire Prompt → Sampling via Connector Edge.

6. Run Validation, fix any errors.

7. Download your manifest.json or Publish.

---

With Manifest Studio, you never write raw JSON—just drag, drop, configure, connect, and export. 🎉