Connect an agent

Availability

Open core · self-host + all Splyntra Cloud plans

The Connect wizard (reachable from Agents → New, or the /connect shortcut) is the fastest way to instrument an agent. It collects a little about your stack, mints an ingest key, and generates connect code you can paste in.

The wizard steps

Step	What you set
1. Basics	The agent name — becomes the service_name / serviceName on its spans.
2. Frameworks	The frameworks you use (LangGraph, CrewAI, OpenAI Agents, …) → the instrument list.
3. LLM Providers	The providers you call (OpenAI, Anthropic, Ollama, or an OpenAI-compatible one) → adds provider adapters and any base_url notes.
4. Data & Tools	Tools, MCP servers, and retrieval sources the agent touches.
5. Security	The inline guard mode — a radio of Off / Monitor / Block (default Monitor).
6. Alerts	A checkbox (default on) to create a HIGH/CRITICAL alert rule for this agent.
7. Review	Confirms your choices, mints the ingest key, and shows the connect code.

Security: guard modes

Mode	Behavior
Off	No inline guard; the pipeline still redacts and scores risk.
Monitor	Runs the pre-flight guard and logs its verdict without altering the call.
Block	Raises SplyntraBlocked before the provider call on an injection or secret hit.

See Guardrails for the full guard behavior.

The ingest key — shown once

On Review, the wizard mints an ingest key and displays it once. Copy it now — it won't be shown again. Keys are stored server-side only as SHA-256 hashes, so a lost key can't be recovered; mint a new one instead. Pass it through an environment variable rather than committing it.

The generated connect code

Review shows a Connect code block with a Python / TypeScript toggle and a Copy button. The snippet sets service_name / serviceName to the agent name, injects the key, applies your chosen instrument list and guard mode, and includes base_url notes for any OpenAI-compatible provider.

Python

import os
from splyntra import Splyntra

Splyntra(
    api_key=os.environ["SPLYNTRA_API_KEY"],
    project="support-agent",
    service_name="support-agent",       # = the agent name from Basics
    endpoint="https://ingest.splyntra.com",
    instrument=("langgraph", "openai"), # from Frameworks + LLM Providers
    guard="monitor",                    # from Security
)

TypeScript

import { Splyntra } from "@splyntra/sdk";

new Splyntra({
  apiKey: process.env.SPLYNTRA_API_KEY!,
  project: "support-agent",
  serviceName: "support-agent",        // = the agent name from Basics
  endpoint: "https://ingest.splyntra.com",
  instrument: ["langgraph", "openai"], // from Frameworks + LLM Providers
  guard: "monitor",                    // from Security
});

Run your agent once, then open Traces to see the first run land, or the Agents screen to see the agent register.

Next steps

• Framework integrations — instrument names and span mapping.
• LLM providers — OpenAI-compatible base_url setup.
• Alerts — tune the HIGH/CRITICAL rule the wizard created.