Strata Documentation
Strata is an AI-empowered document editing platform. Upload Markdown or HTML documents for viewing, commenting, and editing with a section-based document model designed for granular AI editing via MCP (Model Context Protocol).
Quick Start
Connect your AI client to Strata's MCP server to read, edit, search, and manage documents. Most clients handle OAuth automatically — just provide the server URL.
Connection Details
MCP Server URL:
https://api.prod.us-east-2.strata.space/mcpAuthentication: OAuth 2.1 with Dynamic Client Registration
Available Tools:
app_get_section_content,browse_connector_resources,edit_document,find,get_agent_status,get_document_graph,get_image,get_presence,get_publish_status,invoke_agent,invoke_connector_action,list_connected_tools,manage_comments,manage_suggestions,publish_document,read_document,unpublish_document
Client Setup
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"strata": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://api.prod.us-east-2.strata.space/mcp"
]
}
}
}
OAuth authentication is handled automatically — you'll be prompted to sign in on first use.
Claude Code
Add the Strata MCP server via the CLI:
claude mcp add strata https://api.prod.us-east-2.strata.space/mcp
OAuth authentication is handled automatically via your browser.
Cursor
Add to ~/.cursor/mcp.json or .cursor/mcp.json:
{
"mcpServers": {
"strata": {
"url": "https://api.prod.us-east-2.strata.space/mcp"
}
}
}
Cursor handles OAuth automatically when the server returns 401.
VS Code (Copilot)
Add to .vscode/mcp.json in your project:
{
"servers": {
"strata": {
"type": "http",
"url": "https://api.prod.us-east-2.strata.space/mcp"
}
}
}
Requires VS Code 1.101+. Uses "servers" (not "mcpServers") and type "http". OAuth with PKCE and Dynamic Client Registration is handled automatically.
Windsurf
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"strata": {
"serverUrl": "https://api.prod.us-east-2.strata.space/mcp"
}
}
}
Windsurf uses serverUrl instead of url. OAuth is handled automatically.
Cline
Open the MCP Servers panel in Cline and add to the config:
{
"mcpServers": {
"strata": {
"url": "https://api.prod.us-east-2.strata.space/mcp",
"type": "streamableHttp"
}
}
}
Uses "streamableHttp" (camelCase). When OAuth is required, Cline shows an Authenticate button.
Continue
Add to ~/.continue/config.yaml:
mcpServers:
- name: strata
command: npx
args:
- "-y"
- "mcp-remote"
- "https://api.prod.us-east-2.strata.space/mcp"
Continue does not support OAuth natively yet. Use the mcp-remote bridge instead (see below).
Zed
Add to your Zed settings.json:
{
"context_servers": {
"strata": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://api.prod.us-east-2.strata.space/mcp"
]
}
}
}
Zed does not support OAuth natively. Uses mcp-remote as a stdio bridge that handles the OAuth flow in your browser.
Claude.ai and ChatGPT
These chat products render Strata's editor inline as a custom connector. Add the MCP server URL above in the host's connector settings.
Claude.ai
Settings → Connectors → Add custom connector
Available on paid plans. Organization connectors are added by an Owner.
ChatGPT
Settings → Connectors → Create
Requires Developer Mode (Settings → Apps & Connectors → Advanced). Plus, Pro, or Enterprise.
Universal Fallback (mcp-remote)
For any client without native OAuth support, use mcp-remote as a stdio bridge. It handles the full OAuth flow and works with any MCP client:
{
"mcpServers": {
"strata": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://api.prod.us-east-2.strata.space/mcp"
]
}
}
}
Auth state is persisted in ~/.mcp-auth/ so you only authenticate once per server.
Next Steps
See the MCP Tools Reference for complete documentation of all available tools and their parameters.
Agent frontmatter
An agent definition is a Markdown document under your /Agents folder. The YAML frontmatter declares the agent's identity, tool allowlist, and orchestrator visibility; the body below the frontmatter is the system prompt the orchestrator hands the user's request to verbatim. Every field is validated server-side on save.
| Field | Type | Required | Default | Description |
|---|---|---|---|---|
autoInvokable |
boolean | Optional | false | When true, the chat orchestrator may pick this agent on its own if its description matches the user's request. When false (default), the agent runs only on explicit invocation (@mention, MCP invoke_agent, or a connected source). |
description |
string | Required | — | One-sentence summary of when to use this agent. Surfaces in the orchestrator's auto-invoke catalog and the @mention picker — be specific about the agent's job. |
enabled |
boolean | Optional | true | Master switch. When false, the agent is hidden from every invocation surface even when its frontmatter is otherwise valid. |
model |
string | Optional | — | Optional override for the model the agent runs on. Falls back to the platform default when omitted. Must resolve against the platform model registry. |
name |
string | Required | — | Kebab-case identifier ([a-z0-9-]), unique within your agents and not colliding with a reserved platform agent name. Drives the @mention token in chat and the agentName argument to MCP invoke_agent. |
tools |
string[] | Optional | [] | Allowlist of platform tools the agent may call (e.g. read_document, search_space). The Available Tools panel inside the agent banner lists every valid name. Omit or leave empty to grant no tools. |
Example
---
name: meeting-notes-summarizer
description: Summarizes meeting notes into a TL;DR with action items.
tools:
- read_document
- search_space
model: claude-sonnet-4-6
color: emerald
enabled: true
autoInvokable: false
---
You are a meeting-notes summarizer. Given the document body the
orchestrator hands you verbatim, produce a one-paragraph TL;DR and a
bulleted action-item list…
Prompt frontmatter
A prompt template is a Markdown document under your /Prompts folder. The YAML frontmatter declares the prompt's name, description, arguments, and orchestrator visibility; the body below the frontmatter is the prompt template — {{argument}} placeholders are substituted at render time. Argument values are collected by the chat composer's slash sheet or the MCP prompts/get request.
| Field | Type | Required | Default | Description |
|---|---|---|---|---|
arguments |
PromptArgument[] | Optional | [] | Ordered list of PromptArgument entries the prompt body references via {{name}} placeholders. Order is preserved across the wire so the slash-picker renders fields in the author's chosen order. Maximum 16 entries. |
autoInvokable |
boolean | Optional | false | When true, the chat orchestrator may pick this prompt on its own when the description matches the user's intent. When false (default), the prompt runs only when the user types its slash token or a client invokes MCP prompts/get. |
description |
string | Required | — | One-sentence summary of what the prompt does. Surfaces in the slash menu, MCP prompts/list, and (when autoInvokable: true) the orchestrator's tool catalog. |
name |
string | Required | — | Human-readable prompt title. The slugified form becomes the /prompt: |
Example
---
name: Summarize document
description: Summarize the active document for a chosen audience.
autoInvokable: false
arguments:
- name: focus
description: What the assistant should focus on.
required: true
- name: audience
description: Target audience for the summary.
required: false
---
Summarize this document for {{audience}}, focusing on {{focus}}. Keep
the summary under 200 words and finish with a short action-items list.