What Can Claude Code Do? Capabilities Guide for Indian Teams

Claude Code is Anthropic's agentic coding tool for the terminal and IDE. It reads files and PDFs, searches the web, runs commands, edits across a codebase, and connects to internal systems through Model Context Protocol (MCP) servers. It runs on Anthropic's Claude models and is available to Indian enterprises directly or through Amazon Bedrock and Google Vertex AI, which matters for data residency and DPDP Act compliance.

Why Indian teams are evaluating Claude Code

India's GCCs, IT services firms, and product startups run large engineering organisations where consistent throughput and code quality directly affect margins. Claude Code fits that context as an agent that works inside the codebase rather than a chat assistant. For background, see our overview of what Claude Code is for enterprises.

For BFSI and regulated ITeS workloads, the evaluation is rarely just about features. It is about where data is processed and whether the deployment satisfies RBI and CERT-In expectations. Capability and governance have to be assessed together.

Reading documents and accessing the web

Can Claude Code read PDFs?

Yes. Claude Code reads PDFs, images, logs, and source files using the Claude models' vision and document understanding. An Indian fintech team can point it at an NPCI or UPI integration specification in PDF form and have the agent extract the relevant fields before writing the integration code. Where a PDF contains customer or financial data classified as personal data under the DPDP Act 2023, treat that input the same as any other regulated data flow and route it through a compliant deployment.

Can Claude Code search the web?

Yes, through tools you enable. It can look up current framework documentation or verify an API contract before generating code. For CERT-In-aligned environments, web access is controllable, so you can permit it in development and disable it in pipelines that touch production or sensitive data.

Can Claude Code access the internet?

Only when permitted. All network access runs through explicit tools or MCP servers, and administrators can restrict outbound calls. For banks and NBFCs operating under RBI guidance, this lets security teams keep the agent inside an approved network boundary.

Which model powers Claude Code?

Which model does Claude Code use?

Claude Code uses Anthropic's Claude models from the Opus, Sonnet, and Haiku family. Opus-class models suit complex reasoning and large migrations common in GCC modernisation work, Sonnet-class models cover day-to-day engineering, and Haiku-class models handle fast, low-cost tasks. Choose by tier, since Anthropic releases new versions regularly.

Can Claude Code use other models or local models?

No. Claude Code is built for Claude models and is not a bring-any-model or local-model runner. For Indian enterprises, the important point is the deployment path: the same Claude models are available through Amazon Bedrock and Google Vertex AI. Running Claude Code against a Bedrock or Vertex AI account in an Indian or otherwise approved region lets you keep processing within a controlled environment and address DPDP data-residency and RBI localisation requirements through your cloud provider's controls.

How Do You Build software with Claude Code?

Can Claude Code build an app?

Yes, as an engineering collaborator. It can scaffold services, implement features across files, write and run tests, and debug until the build passes. For Indian product startups, that means faster delivery on a lean team; for ITeS firms, more predictable output on client engagements. It is not a no-code generator. It produces real code that your engineers review and own, which keeps accountability and IP clearly with your organisation, an important point in client and audit conversations.

Can Claude Code generate images?

Can Claude Code generate images?

Not on its own. Claude Code is a coding agent, not an image model. It does not create illustrations or graphics natively. Through MCP you can attach an image-generation service as a tool and have the agent invoke it within a workflow, but the generation runs in that external tool. For most enterprise engineering use cases this is not a requirement.

Capability summary

Connecting internal systems with MCP

Where is the Claude Code MCP config?

MCP servers are defined in Claude Code's configuration files, typically a .mcp.json at the project level alongside user and enterprise scopes. A GCC platform team can commit a shared MCP configuration to the repository so every engineer connects to the same internal APIs, databases, or ticketing systems with consistent, reviewed access. Because access is scoped and file-based, security teams can audit exactly which internal systems the agent can reach, which supports CERT-In-aligned change and access governance.

Will Claude Code replace developers?

Will Claude Code replace developers?

No. For India's services and GCC model, the realistic outcome is higher output per engineer and a shift toward design, review, and quality, not workforce replacement. Claude Code absorbs boilerplate, test scaffolding, and routine refactors, while engineers own architecture, security, and client commitments. Human review remains mandatory, especially in BFSI where correctness is regulated. For a tooling comparison, see Claude Code vs Cursor.

Frequently asked questions

Does Claude Code support data residency for Indian compliance?

Indirectly, through deployment. Running Claude Code via Amazon Bedrock or Google Vertex AI in an approved region lets you apply your cloud provider's residency and security controls, which is the usual route for DPDP and RBI-aligned workloads.

Can it run in our CI/CD pipelines?

Yes. It integrates with GitHub Actions and similar pipelines for automated reviews and changes. See our guide to Claude Code with GitHub Actions.

Can we restrict what commands it runs?

Yes. Permissions are configurable: allow safe operations, block destructive ones, require approval, or run read-only agents that analyse without modifying code.

Does it work on large legacy codebases?

Yes. It is built for real repositories and uses search and context management to operate across many files, which suits the modernisation work common in Indian GCCs and IT services.

Read more about gdpr compliance from Opsio.