Agents

List and inspect agents, and run the local code dev loop — pull, push, deploy.

Agents are the steps your pipeline runs. Premia ships built-in agents (planner, developer, reviewer, QA, …) and lets you add custom agents — either config-only or full code agents you edit locally and deploy to our cloud.

The code dev loop is: pull → edit → push → deploy.

The run(context) contract

Code agents export a single run(context) function. Premia wraps it in a hosted managed function (image, secrets, logging) — you never manage infrastructure. See Custom code agents for the full contract.

`premia agents list`

List every agent available in your org (built-in and custom).

premia agents list [--json]

SLUG            NAME              TYPE      RUNTIME  DEPLOY
plan            Planner           built-in
execute         Developer         built-in
review          Reviewer          built-in
security-lens   Security Lens     code      python   deployed
triage-bot      Triage Bot        config

TYPE is built-in for Premia's agents, otherwise the custom agent type (code or config). DEPLOY shows the deployment status for code agents.

`premia agents view`

Show one agent's configuration.

premia agents view <slug> [--json]

Security Lens (security-lens)
Reviews diffs for auth and data-exposure issues.
type    code
model   claude-sonnet-4-6
runtime python
deploy  deployed

`premia agents pull`

Download a code agent's source to a local file so you can edit it.

premia agents pull <slug> [--force]

Flag	Description
`--force`	Overwrite an existing local file.

Writes ./<slug>.py (or .ts for TypeScript agents). If the agent has no code yet, Premia writes a starter template so the contract is obvious:

security-lens.py

# Custom Premia agent. You write run(context); Premia wraps it in a managed
# function (image, secrets, logging) and hosts it — no infra to manage.
#
# context holds the data you bound to this agent (run, repo, pr_diff, plan_output, …).
# 'logger' is available (logger.log("info", "…")). Return a dict.

def run(context: dict) -> dict:
    logger.log("info", "hello from my agent")
    return {"status": "done", "output": "edit me"}

✓ Pulled security-lens → security-lens.py

`premia agents push`

Upload your local changes back to the agent (this stages the code; it goes live on the next deploy).

premia agents push <slug> [--file <path>]

Flag	Description
`--file <path>`	Path to the source file. Defaults to `./<slug>.<ext>`.

✓ Pushed security-lens.py → security-lens (1284 bytes)
  Deploy it: premia agents deploy security-lens

If the default file isn't found, the command exits 3 and tells you to pass --file or run premia agents pull <slug> first.

`premia agents deploy`

Build and deploy a code agent to our cloud.

premia agents deploy <slug> [--wait]

Flag	Description
`--wait`	Block until the deploy finishes (`deployed` or `failed`).

→ Deploying security-lens…
  Check status: premia agents view security-lens

With --wait, the CLI polls every 4s and reports the outcome, exiting 1 if the build fails:

→ Deploying security-lens…
✓ deployed

The full loop

Pull the agent to a local file: premia agents pull security-lens

Edit security-lens.py — implement your run(context).

Push your changes: premia agents push security-lens

Deploy to the cloud: premia agents deploy security-lens --wait

See Models & secrets for choosing a model and giving your agent credentials.

Agents

premia agents list

premia agents view

premia agents pull

premia agents push