For Claude and Anthropic builders

Observability built for how you actually use Claude.

Morse instruments the Anthropic SDK, the Claude Agent SDK, and the Vercel AI SDK — then turns labeled production outcomes, spend, and separate sampled quality signals into a ranked performance brief, with traces and evidence on demand.

What Morse shows for your Claude agents

Four signals, kept separate on purpose — Morse never blends cost and quality into one number.

Prompt-cache economics, visible.

Anthropic's cache pricing rewards well-structured prompts — Morse estimates cache savings with provider-specific pricing and highlights low hit rates.

Memory instrumentation that understands Claude apps.

Typed memory operations captured through a client wrap — compare their cost and latency, plus directional success rates for runs with and without memory.

Cost per successful outcome (CPSO) — not cost per run.

CPSO divides spend by outcomes your team can actually confirm, not by every run. When nothing is confirmable yet, Morse shows no CPSO instead of a misleading number.

Sampled quality, kept separate.

Automatic scoring and sampling run alongside cost and outcomes but never fuse into one score — each signal keeps its own coverage and provenance so you can tell what's backing a verdict.

Instrument the Anthropic SDK in one line

Wrap your existing Anthropic client. Morse captures streaming, tool use, and prompt-cache tokens directly — no proxy in the request path.

Python

pip install morse-ai

from morse_ai.adapters import anthropic as morse_anthropic
import anthropic

client = morse_anthropic.wrap(anthropic.Anthropic())

# Every messages.create() call now emits an llm span with
# pre-segmented context and prompt-cache token detail.

TypeScript

npm install @morsehq-dev/sdk

import { wrapAnthropic } from "@morsehq-dev/sdk/anthropic";
import Anthropic from "@anthropic-ai/sdk";

const client = wrapAnthropic(new Anthropic());

// Every messages.create() call now emits an llm span with
// pre-segmented context and prompt-cache token detail.

Building with the Claude Agent SDK or the Vercel AI SDK instead? Dedicated adapters — @morsehq-dev/sdk/anthropic-agent-sdk and @morsehq-dev/sdk/vercel-ai — cover those the same way.

When something needs investigating, Morse produces hypotheses with evidence attached — not a claimed root cause. You decide what to trust and what to check yourself.

See what a performance brief looks like for a Claude agent.

Join the waitlist, or look at a sample brief first.

Join the waitlist See a sample brief →