Pricing
Fixed-scope pricing. No surprises.
Every project gets a written quote before work starts. The number in the quote is the number on the invoice. No hourly billing. No change orders without approval. No retainer lock-in.
Starter agent
Timeline: 3 to 5 days
Single-workflow agents with 1 to 3 tools and a clear input/output contract.
- ✓ One agent, one workflow
- ✓ Up to 3 MCP tool integrations
- ✓ Basic Langfuse observability
- ✓ Handover runbook
- ✓ 30-day email support
Examples
- FAQ chatbot on a help doc set
- Lead routing agent
- Simple form-to-CRM workflow
Production agent
Timeline: 1 to 3 weeks
Agents with multi-step workflows, memory, and integrations across 3 to 8 tools.
- ✓ One production-grade agent
- ✓ Up to 8 MCP tool integrations
- ✓ Persistent memory via Mem0
- ✓ Langfuse observability with cost dashboard
- ✓ Eval suite with 50 golden examples
- ✓ CI eval gate before shipping
- ✓ Full codebase handover + CLAUDE.md
- ✓ 30-day Slack support
Examples
- Voice receptionist (Retell + Sarvam + Cal.com)
- RAG customer support (Mastra + Qdrant + Voyage AI)
- AI SDR with human approval gate
- WhatsApp business agent for India
Multi-agent system
Timeline: 3 to 6 weeks
Complex orchestrated systems with supervisor agents, parallel workers, and shared state.
- ✓ Multi-agent graph (LangGraph or Mastra)
- ✓ Supervisor + worker architecture
- ✓ Shared memory with conflict resolution
- ✓ Per-agent eval suites
- ✓ Full observability: traces, cost attribution, latency
- ✓ Staging environment + production cutover
- ✓ Team training session (2 hours)
- ✓ 60-day Slack support
Examples
- SaaS onboarding pipeline (48x faster)
- AI coding team setup for engineering orgs
- Multi-step research and content agent
Optional retainer
Monthly maintenance
Agents drift. Models update. Prompt patterns change. The retainer keeps your agent accurate and cost-efficient without requiring you to manage it yourself.
- ✓ Monthly prompt audits and adjustments
- ✓ Token budget monitoring and alerts
- ✓ Model updates when new Claude versions change behavior
- ✓ Up to 5 hours of changes per month (new intents, integrations, policy updates)
- ✓ Priority response time: 4 hours vs 24 hours
Typical operating costs
These are billed directly to your accounts, not through us. We set up monitoring and alerts for all of them.
| Service | Cost |
|---|---|
| Claude Sonnet 4.6 (input) | $3 / 1M tokens |
| Claude Sonnet 4.6 (output) | $15 / 1M tokens |
| Claude Haiku 4.5 (input) | $0.80 / 1M tokens |
| Retell AI voice | $0.07 / minute |
| Sarvam voice (India) | $0.04 / minute |
| WhatsApp marketing message (India) | ₹0.86 + 18% GST |
| Qdrant Cloud (hosted) | From $25 / month |
| Mem0 memory layer | From $0 (OSS self-host) |
| Upstash Redis (rate limiting) | From $0 (free tier) |
| Langfuse observability | From $0 (OSS self-host) |
Pricing questions
- Are these fixed prices or estimates?
- Every project gets a fixed-scope quote before work starts. The price in the quote is the price you pay. We do not bill hourly and we do not have change-order surprises. If scope changes, we send a revised quote and wait for approval.
- Do you bill for model API usage?
- No. Model API keys live in your account, billed directly to you. The operating cost table on this page shows typical costs per usage unit. We set up cost alerts and monthly budget caps as part of every build.
- What about INR pricing for Indian clients?
- We quote in both USD and INR. INR pricing is converted at the current rate and rounded to the nearest 5K. For Indian clients we accept bank transfer, Razorpay, and UPI for amounts under ₹5L.
- Is there a discovery fee?
- No. The first 30-minute call is free. After that call, we send a scope document at no charge. You decide whether to proceed. If you do not, you owe nothing.
Ready to ship something real?
Book a 30-minute build call. Walk away with a plan either way.