Skip to content
cccskills
FreePublished on GitHub

anth-deploy-integration

Prepare Claude API services for deployment using Docker, Cloud Run, or Kubernetes. Configure runtime secrets, health checks, staging tests, gradual releases, and rollback procedures, then document deployment evidence without exposing credentials or request content.

OriginalShow the original description

Deploy Claude API integrations to production cloud environments. Use when deploying Claude-powered services to Docker, Cloud Run, ECS, or Kubernetes with proper secret management and health checks. Trigger with phrases like "deploy anthropic", "claude production deploy", "ship claude integration", "anthropic cloud deployment".

Run score Not tested

See how to install

Good for

  • Deploy a Claude API service to Cloud Run
  • Configure Kubernetes secrets and health checks
  • Validate staging before a limited release
  • Prepare rollback procedures for failed releases

Files included (1)

  • SKILL.md4.9 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Anthropic Deploy Integration

Overview

Deploy Claude API integrations with proper secret management, health checks, and rollback procedures across Docker, GCP Cloud Run, and Kubernetes.

Docker Deployment

FROM python:3.12-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY src/ ./src/
ENV ANTHROPIC_API_KEY=""
EXPOSE 8000
CMD ["uvicorn", "src.main:app", "--host", "0.0.0.0", "--port", "8000"]
# src/main.py
from fastapi import FastAPI, HTTPException
import anthropic

app = FastAPI()
client = anthropic.Anthropic()

@app.get("/health")
async def health():
    try:
        count = client.messages.count_tokens(
            model="claude-haiku-4-20250514",
            messages=[{"role": "user", "content": "ping"}]
        )
        return {"status": "healthy", "api": "connected"}
    except Exception as e:
        raise HTTPException(503, detail=str(e))

GCP Cloud Run

echo -n "sk-ant-api03-..." | gcloud secrets create anthropic-key --data-file=-

gcloud run deploy claude-service \
  --image gcr.io/my-project/claude-service \
  --set-secrets ANTHROPIC_API_KEY=anthropic-key:latest \
  --min-instances 1 --max-instances 10 \
  --memory 512Mi --timeout 120s

Kubernetes

apiVersion: apps/v1
kind: Deployment
metadata: { name: claude-service }
spec:
  replicas: 3
  strategy: { type: RollingUpdate, rollingUpdate: { maxUnavailable: 1 } }
  template:
    spec:
      containers:
        - name: app
          env:
            - name: ANTHROPIC_API_KEY
              valueFrom:
                secretKeyRef: { name: anthropic-secrets, key: api-key }
          livenessProbe:
            httpGet: { path: /health, port: 8000 }
            periodSeconds: 30

Rollback

# Cloud Run
gcloud run services update-traffic claude-service --to-revisions=PREVIOUS=100

# Kubernetes
kubectl rollout undo deployment/claude-service

Error Handling

IssueCauseFix
Container crash on startMissing API key env varVerify secret binding
Health check failsKey invalid in prodTest key with curl
429 after scaling upMore replicas = more RPMShared rate limiter (Redis)

Prerequisites

  • Have an approved artifact digest, environment/workspace mapping, secret-manager reference, health probe, deployment owner, canary plan, and tested rollback command.
  • Use a least-privileged runtime identity and synthetic fixtures in staging; never embed API keys in images, manifests, command history, or deployment output.
  • Define deployment SLOs for health, errors, latency, rate limits, cost, and data-policy checks, with explicit halt thresholds.

Instructions

  1. Build and scan the pinned artifact, bind the environment-specific secret at runtime, and verify that logs and probes cannot expose prompts, responses, or credentials.
  2. Deploy to staging and run the token-count/health probe plus synthetic Messages API, error, timeout, rate-limit, and redaction tests. Confirm the workspace and model are approved.
  3. Release to one sandbox or internal canary, monitor aggregate SLOs and sensitive_content_logged=0, and require owner approval before wider traffic.
  4. Promote in bounded stages while preserving the prior revision and idempotent deployment record. Do not bypass failed health, authorization, or data-policy gates.
  5. On failure, stop traffic, roll back to the prior revision, revoke temporary access, clean up staged artifacts under the retention policy, and issue a redacted deployment receipt.

Output

Produce a deployment receipt containing artifact digest, environment/workspace class, model policy, probe/test results, canary scope, SLO outcomes, approval, rollout state, retention cleanup, and rollback reference. Exclude API keys, prompts, responses, customer identifiers, and raw stack traces.

Examples

Deploy artifact=sha256:fixture to a staging workspace, run synthetic fixture-request-001, assert workspace=staging; sensitive_content_logged=0, then release a 1% internal canary. If the 5xx or latency gate fails, record promotion=halted; rollback=previous-revision and send no production traffic.

Resources

Next Steps

For event-driven patterns, see anth-webhooks-events.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Add a synthetic-data CI gate for customer-owned Abridge integration code. Test consent handling, note-review handoffs, redaction, and rollback behavior with mocked vendor boundaries, without connecting CI to a live Abridge or EHR tenant.

  • Adding regression checks to an Abridge integration repository
  • Testing consent and note-release handling
  • Checking logs for sensitive data exposure
jeremylongshore/tons-of-skills-marketplace2.9KUpdated Oct 25, 2026

Triage Abridge recording, note review, Epic handoff, and access problems using observed symptoms and approved guidance. Identify the affected workflow stage, organize a minimal timeline, and prepare a redacted escalation record for the responsible team.

  • Separate recording problems from upload failures
  • Investigate note creation delays and editor issues
  • Identify the responsible team for Epic handoff failures
jeremylongshore/tons-of-skills-marketplace2.9KUpdated Oct 25, 2026

Map and review an Abridge clinical documentation workflow from consent and recording through draft review, evidence checks, edits, and final disposition. Keep clinician approval explicit, identify failure exits, and separate verified facts from tenant-specific details and pending approvals.

  • Map a clinical note workflow for a care setting
  • Check where consent and patient identity are verified
  • Review draft notes against linked evidence before sign-off
jeremylongshore/tons-of-skills-marketplace2.9KUpdated Oct 25, 2026

Audit how clinician-reviewed Abridge note sections map into Epic templates and downstream workflows. Use approved test records to document destinations, exceptions, evidence, rollback steps, and a go or no-go decision for a specific environment.

  • Checking Abridge note section mappings to Epic templates
  • Auditing a department’s approved documentation handoff
  • Preparing test evidence and a go or no-go decision
jeremylongshore/tons-of-skills-marketplace2.9KUpdated Oct 25, 2026

Review Abridge adoption and workflow value using local usage records, clinician feedback, and verified contract costs. Build an evidence based decision table covering adoption, quality, support burden, and options to expand, hold, or improve the rollout.

  • Reviewing adoption after an Abridge rollout
  • Preparing evidence for a renewal decision
  • Comparing workflow outcomes across care settings
jeremylongshore/tons-of-skills-marketplace2.9KUpdated Oct 25, 2026

Prepare a bounded support bundle for an Abridge issue using allowlisted operational details such as versions, coarse timestamps, and correlation references. Exclude recordings, transcripts, patient identifiers, credentials, and unrestricted logs, then document redactions and review details.

  • Preparing reviewed evidence for an Abridge support case
  • Documenting redactions and excluded data
  • Collecting approved operational details for an incident
jeremylongshore/tons-of-skills-marketplace2.9KUpdated Oct 25, 2026

See all skills by jeremylongshore

Report a problem with this skill