Developer integrations

Use Architect inside another app.

Send Architect a rough employee prompt, let it use the company's knowledge files, then return an enriched prompt for Claude Code, Codex, or an internal AI.

Upload company knowledge

Add PDFs or text files in Memories. Enabled files are retrieved automatically by API calls.

Create an API key

Create keys in Settings. Keys are shown once and must be sent as a Bearer token.

Call HTTP, MCP, or compiler

Use enrichment for AI-powered rewriting and the compiler for deterministic diagnostics and handoff packaging.

Direct HTTP endpoint

curl -X POST https://architect-ai.online/api/integrations/enrich-prompt \
  -H "Authorization: Bearer arch_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "request": "make this ticket into a precise implementation prompt",
    "category": "Coding",
    "language": "en",
    "targetTool": "Codex",
    "handoffFormat": "codex",
    "conversationContext": "User is asking for a bug fix in checkout.",
    "requireQuestions": true
  }'

Company app pipeline

const architectResponse = await fetch("https://architect-ai.online/api/integrations/enrich-prompt", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.ARCHITECT_API_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    request: workerMessage,
    category: "Business",
    targetTool: "Internal company AI",
    handoffFormat: "internal_ai",
    downstreamContext: JSON.stringify({ department, currentWorkflow }),
    responseMode: "text"
  })
});

const enrichedPrompt = await architectResponse.text();
const internalAiResult = await yourInternalAi.generate(enrichedPrompt);

Handoff formats

general: for any assistant.
claude_code: for Claude Code tasks with repo context and verification steps.
codex: for Codex coding work with safe edit boundaries.
internal_ai: for a company assistant that must follow uploaded company context.

Manage knowledge Create API key

MCP-compatible call

POST /api/mcp
Authorization: Bearer arch_YOUR_KEY

{
  "jsonrpc": "2.0",
  "id": "1",
  "method": "tools/call",
  "params": {
    "name": "architect_enrich_prompt",
    "arguments": {
      "request": "turn this rough idea into a safe coding task",
      "targetTool": "Claude Code",
      "handoffFormat": "claude_code",
      "language": "en"
    }
  }
}

Prompt compiler endpoint

curl -X POST https://architect-ai.online/api/integrations/compile-prompt \
  -H "Authorization: Bearer arch_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "roughPrompt": "make a prompt for a clothes website",
    "finalPrompt": "Role: You are... Objective: Create...",
    "category": "Business",
    "targetTool": "codex"
  }'

Compiler use case

Run this after Architect or another app creates a prompt. It returns score, risks, missing decisions, test cases, and a version formatted for Codex, Claude Code, general assistants, or internal AI.

Open Prompt Lab