Large requests go wrong when Claude starts coding before the scope is clear. The tech lead sits in front of the other specialists: it decides whether a task needs a quick fix or a full plan, and hands each part to the right agent with acceptance criteria attached.
When to use the Tech Lead agent
- A feature that touches several parts of the codebase
- Work in a repository you or Claude have not seen before
- Architecture decisions with trade-offs to record
- Anything that needs more than one specialist
What is in the agent file
| Name | tech-lead |
|---|---|
| Tools | Read, Grep, Glob, Bash, TodoWrite. No write or edit tools, so it reports and does not change files. |
| Model | opus, up to 50 turns per task |
| Kit | Engineer kit, 123 lines |
How it works. Beyond its role statement, the file sets out 4 rules it follows: Workflow selection; Risk classification: drives how much process; Delegation brief format; Completion evidence.
What it reports. Every run ends in the same structure: Objective, Workflow chosen, Plan, Decisions, Status, Risks, Approvals needed, Next action.
Excerpt from the agent file
The settings and opening instructions, as they appear in .claude/agents/tech-lead.md:
---
name: tech-lead
description: Orchestrates complex software work from request through verified delivery — choosing the smallest sufficient workflow, briefing specialists, and keeping risk and scope visible. Use PROACTIVELY for cross-cutting features, unfamiliar repositories, architecture decisions, or work needing multiple specialists.
model: opus
maxTurns: 50
tools: Read, Grep, Glob, Bash, TodoWrite
---
You are a Tech Lead. Your job is to get work delivered and verified, choosing the least process that will do it.
When given a request:
1. Clarify the objective and name what is missing. Do not start work on an ambiguous request.
2. Inspect repository conventions and available tooling before deciding anything.
3. Choose the workflow from the table below. Over-process is as costly as under-process.
4. Produce a brief per specialist — context, scope, acceptance criteria, constraints. You cannot spawn subagents, so name the specialist and hand the caller a brief they can dispatch directly.
5. Review plans before implementation begins.
6. Track scope, risk, dependencies and decisions as they change.
7. Require tests and evidence before reporting anything complete.The full file continues with the rules above and the exact report format. It is included in the Agentsforce kit.
Example requests
Name the agent to be sure Claude uses it, or describe the task and let its description trigger it:
- "Use tech-lead to plan adding team accounts to this app. Don't write code yet."
- "We need to move from REST to webhooks for billing. Have tech-lead scope it and list the risks."
- "Use tech-lead on this feature request and give me the briefs for each specialist."
- "Tech-lead: is this a one-day change or a migration? Tell me what process it needs."
When to use a different agent
| If you need | Use instead |
|---|---|
| Writing the code itself | It plans and briefs; senior-implementer or the main session builds |
| A one-line fix | Its own rules say over-process costs as much as under-process; just ask the main session |
Works well with
- Code Reviewer: checks what a change did, not whether a file looks fine
- API Engineer: designs and changes endpoints that clients depend on
- Database Engineer: changes schemas and queries without risking your data
Related playbooks in the kit, which Claude loads when the task calls for them: acceptance-criteria, vertical-slicing, architecture-decision-records.
How to get it
- Write your own: the free Subagent Generator drafts a starter tech lead file you can edit and save in
.claude/agents/. - Use the tested version: Agentsforce includes this agent with 54 other specialists and 89 playbooks for a one-time $49, installed with one command.
Frequently asked questions
Can the tech lead agent start other subagents?
No. Subagents cannot spawn other subagents in Claude Code, so it names the specialist and writes a brief you or the main session can dispatch directly.
Why does it run on Opus?
Planning across a whole codebase involves more ambiguous reasoning than a single scoped task, so its file sets the model to opus. You can change the model line if cost matters more.
Does it edit files?
No. Its tools are Read, Grep, Glob, Bash and TodoWrite, so it can inspect the repository and track tasks but not change code.