Give your coding agent a real workspace: Tools, tests, Git & MCP
Giving your coding agent an environment to act in
An agent is only as good as the tools it can act through. In this session, Lead AI Architect Kwasi Ankomah gives the series’ coding agent a real workspace — files, code execution, tests, and Git — sandboxes it safely, and extends it with developer tools over MCP.
About this webinar
A coding agent that can only talk about code isn’t much use. To do real work, it needs to read and write files, run what it writes, check its changes against a test suite, and manage its work in version control — all without putting your systems at risk.
In Part 3 of the SambaNova coding agent series, Kwasi Ankomah turns the agent from a planner into a doer. He builds out a real development environment for it, isolates it in a safe sandbox, and connects it to developer tools through the Model Context Protocol (MCP). By the end of the session, the agent opens a PR-style diff that passes CI.
What you will learn
- How to give a coding agent a real workspace: reading and writing files, running code, and executing the test suite
- How to let the agent work with Git so its changes are tracked, reviewable, and reversible
- How to sandbox the agent so it can act freely without touching anything it shouldn’t
- How to extend the agent with developer tools over MCP
- How to take the agent all the way to a PR-style diff that passes CI
Why the agent environment matters
Most of a coding agent’s capability comes from what it can actually do, not what model it runs on. Without access to files, tests, and version control, an agent can only suggest code and hope it works. With the right tools and a safe sandbox, it can verify its own changes, recover from errors, and produce work ready for review. MCP makes this extensible, so teams can plug in new developer tools without rebuilding the agent.
Who should attend
AI and ML engineers building coding agents or developer tooling, software engineers exploring agentic workflows, platform and DevOps engineers responsible for running agents safely, and technical leads evaluating MCP and agent sandboxing. Following earlier parts of the series is helpful but not required.
Kwasi Ankomah is the Lead AI Architect at SambaNova Systems, where he builds production agentic AI systems on purpose-built RDU inference infrastructure. With 15 years of experience spanning financial services, consulting, government, and tech startups, Kwasi bridges deep technical expertise with a hands-on understanding of real business challenges. He specializes in deep agent architectures, multi-agent orchestration, and context engineering. Connect with Kwasi on LinkedIn →