Quickstart

Get from zero to a working orchestrated call in about ten minutes. This path uses the orchestration runner template so you can trigger Prscnt from any UI. If you just want to try the tools by hand first, connect the MCP in Claude (Option B below) and talk to it.

1. Get a key

Connect Prscnt to Claude (one click, no key) at mcp.prscnt.com/connect — that auto-provisions your free workspace. For a headless runner that can't open a browser, ask Prscnt to mint an API key once you're connected. Your workspace is resolved from that key — you never pass a workspace id. See Connect & auth.

2. Create the agent runner

The runner is a tiny webhook server that holds the Prscnt MCP connection and runs an LLM agent loop. The full template — five short files — is on The orchestration runner page. Make a new folder, copy those files in, then:

cp .env.example .env        # set ANTHROPIC_API_KEY and PRSCNT_API_KEY
npm install
npm run dev                 # webhook server on :8787

You need two keys: your Anthropic key (the LLM that runs the loop) and your Prscnt key (the intelligence + pipeline). Both are yours; the agent runs on your side. Nothing to clone.

3. Make your first orchestrated call

curl -s localhost:8787/run \
  -H 'content-type: application/json' \
  -d '{"task":"What is the going rate for a TikTok reel from a mid-tier creator with the brand Kerrygold?","context":{}}'

{ "ok": true, "result": "For a mid-tier creator, a single TikTok reel with Kerrygold benchmarks around ... (recommended / floor / ceiling), based on closed-deal data." }

That round trip is the whole pattern: your request went to the runner, an LLM picked the right Prscnt tools (here brand_resolve then rate_benchmark), and you got a plain-language answer back. Swap the curl for a button in your UI and you're integrated.

4. Trigger it from your UI

The runner is just HTTP, so any UI can call it. The trigger differs by product; the call is identical:

Your UI How it fires POST /run
Softr Button → automation → "Make an API request"
Retool A REST query resource pointed at the runner
Airtable Automation → "Run a script" or a webhook step
Custom app A server route that POSTs to the runner
Cron / queue A scheduled job that POSTs a task

Pass whatever your UI knows as context (a record id, a deal id, a brand name). The agent uses it to do the work and — once you wire your own data connector — writes results back. See Compose your stack.

Alternative: connect in Claude (no code)

If your team already works in Claude, skip the runner. In Claude → Settings → Connectors → Add custom connector, paste:

https://mcp.prscnt.com/v1/mcp

Authorize once, and talk to it: "Find brands that fit my roster and draft a pitch to the best one in my voice." Connect your own stack MCPs in the same session and Claude composes them. This is the fastest way to feel what the tools do before you build.

Next

← OverviewCore concepts →