Skip to content
Tembo Agent Studio Tembo Agent Studio

Authoring agents

Agents are authored as files and changed through pull requests. You can write those files directly, but the usual path is to describe what you want and let TAS open the PR for you.

Creating an agent from chat

Describe the agent you want — its job, its tone, the services it should use. TAS hands the request to the Tembo Coding Agent Platform, which writes the agent file and opens a pull request against your repo. Review and merge it; the agent then shows up in the Agents list. (This requires a Tembo API key in Settings.)

The agent file (Pydantic AgentSpec)

A minimal agent is a YAML file under agents/pydantic-agentspec/:

name: standup-summary
model: anthropic:claude-sonnet-4-6
description: Summarize yesterday's commits into a standup note.
instructions: |
  When invoked, summarize the team's activity in three bullet points.
  Be concise and factual.

Key fields:

  • name (required) — must match the filename (name: foo → foo.yaml); lowercase letters, digits, hyphens.
  • model (required) — provider:model, e.g. anthropic:claude-opus-4-8, anthropic:claude-sonnet-4-6, openai:gpt-5.2, openai:gpt-4o-mini. The provider’s key must be set in Settings → API keys.
  • instructions (required) — the system prompt, usually a | block scalar.
  • connections: (optional) — external services the agent calls; see Connections.
  • tools_module: (optional) — a sibling Python file of deterministic tool functions; see Sidecar Python tools.
  • labels: (optional) — tags used for grouping and for scoping which Slack app may launch the agent.

Your connected repo also carries an authoring guide (AGENTS.md and per-framework AGENT_GUIDE.md) that TAS keeps current — that’s the canonical, always-up-to-date field reference for coding agents.

Choosing a model

Model choice is a cost/reliability tradeoff:

  • Start tool-using agents on a top-tier model (e.g. anthropic:claude-opus-4-8). Lower tiers tend to hedge on tool use — asking “would you like me to…” instead of acting — and a decisive model is easier to prove out.
  • Then downgrade and measure. anthropic:claude-sonnet-4-6 is much cheaper and usually fine when the agent has a single, well-defined job with imperative instructions and narrow connections:. Compare cost side-by-side on the Runs page.
  • No tools? Sonnet is a fine starting point — the hedging problem only shows up with tool use.

Iterating

  • Chat with an agent to probe its behavior against the live draft, then submit a change request that opens a PR.
  • Improve the Agent from any run turns feedback into a PR — see Improvements.
  • Promote a draft to a stable version when you’re happy with it; runs serve the stable snapshot by default. See Core concepts → versioning.