Introducing the Giga CLI

Jul 13, 2026

The Giga CLI — a pixel GIGA wordmark over an ambient terminal field, with a CLI prompt cursor

Improving a production support agent usually means a lot of context switching: open the dashboard, find the conversation, figure out what went wrong, edit the config, push, wait for the sync, discover a broken reference, and start over. If you spend your day in a terminal, or your coding agent does, that loop is slower than it needs to be.

Today we’re launching the Giga CLI. It’s a single tool, giga, that gives you the full Giga platform from the command line: agents, tickets, tests, KPIs, secrets, knowledge base, and deploys. Anything you can do in the console, you can now do from your terminal, a script, a CI job, or a coding agent.

pip install giga-sdk
giga login

That’s the whole install. The package ships both the CLI and the Python SDK (from giga import Agent) from one wheel, so the tool you script with and the library you build with stay in lockstep.

~ % pip install giga-sdk
Successfully installed giga-sdk-1.0.0
 
~ % giga login
pip install giga-sdk · giga login

Agents as code

Giga agents are repo-backed. Your agent’s behavior — policies, intents, tests, code blocks — lives in a git repository, and every change is a commit. The CLI is the control plane for that model:

giga validate          # catch broken config before it ships
giga tests run --gate  # run the agent's test suite, non-zero exit on failure
giga git-sync run      # deploy the change
Fix identified · policies/refunds.yaml
 
giga % vim policies/refunds.yaml
 
giga % git diff
 
diff --git a/policies/refunds.yaml b/policies/refunds.yaml
@@ -12 +12 @@
- auto_approve_limit: 0
+ auto_approve_limit: 50
 
giga % git commit -am 'allow auto-approval under $50'
[support-agent 8b4e1d0] allow auto-approval under $50
1 file changed, 1 insertion(+), 1 deletion(-)
Every change is a commit: the fix, the diff, the hash.
Ship and operate your support agents. From the terminal.
• signed in as kyle@giga.ai · giga-sdk 1.0
Run giga --help to get started.
~ % cd giga
giga % giga validate
✓ Successfully validated 1 agent(s), 2 shared policies, 8 s…
giga % giga tickets list -m chat
giga validate — the sub-second local check.

Every change is versioned in git, so history, diffs, and rollback work the way they do for the rest of your production software. No “who changed the agent and when” archaeology in a web UI.

Start from a real conversation

Here’s the loop we use ourselves. A customer escalation comes in:

giga tickets show TICKET-1842 --transcript
giga % giga tickets list -m chat
 
Tickets (chat)
IDConv IDMediumStatusCreated
TICKET-18399f2e4a71b0c4chatuser_ended2026-…
TICKET-1840b6d1c07e52aachatuser_ended2026-…
TICKET-18414c9a8e22f1d7chatuser_ended2026-…
TICKET-1842e3b09c55a2f8chattransferred2026-…
giga %
The queue catches TICKET-1842 — transferred.

You can ask the agent to explain its own behavior:

giga chat ask "Why did this conversation fail?" --agent support-agent
Resolution abandoned
 
giga % giga scout ask --follow "Why did TICKET-1842 fail?"
 
Read    TICKET-1842 · 23 turnsdone · 0.8s
Read    policies/refunds.yamldone · 0.3s
Check  refund flow · 4 policy pathsdone · 1.2s
 
The agent could not process the refund.
Policy requires manual approval for every
refund. No auto-approval path exists.
 
policy gap · refunds.auto_approve_limit: 0
 
Fix identified · policies/refunds.yaml
Scout isolates the policy gap.

Giga traces the conversation, isolates the policy gap, and points at the part of the config that needs to change. Make the fix in the repo, giga validate, giga tests run --gate, deploy, and then check whether it actually moved the number:

giga kpis show resolution-rate
resolution-rate · since deploy
 
82.4%
 
was 78.2% before deploy · up 4.2 pts
Measured impact: resolution-rate, up 4.2 pts since deploy.

Issue to diagnosis to tested fix to measured impact, without leaving the terminal.

Built for pipes, and for agents

Every command supports --format json, so the CLI composes with the tools you already use:

giga tickets list --lucene-query "tag:refund" --format json \
  | jq -r '.[].id'

That same property is what makes the CLI a tool surface for coding agents. Point Claude Code — or any agent that can run shell commands — at a Giga repo and it can inspect tickets, run tests, validate changes, and check KPIs on its own, with structured output it can actually parse. We know this works because it’s how our own agent, Scout, operates the platform.

Claude Code v2.1.207
Fable 5 with xhigh effort · Claude Max
/Users/kyle/giga
>Try "ship the refund policy fix"
? for shortcuts
Claude Code, pointed at a Giga repo.
# scout · autonomous session
 
giga % giga tickets list -m chat \
--lucene-query "status:transferred"
1 ticket · TICKET-1907 · payment_failed
 
giga % giga chat ask -a support-agent \
"my payment failed, can you retry it?"
agent: I cannot retry payments. Escalating to a human.
Scout operates the platform through the same CLI.

Some things this makes possible — starting with the CI gate itself:

82.4%
 
was 78.2% before deploy · up 4.2 pts
giga % cat .github/workflows/giga.yml
 
on: pull_request
 
jobs:
agent:
runs-on: ubuntu-latest
steps:
- run: giga validate --deep
- run: giga tests run --agent support-agent --all
The same gate, encoded in CI.

Operate agents like production software

Enterprise agent deployments should be operated like production software: validated before deploy, tested in CI, versioned in git, observable from the command line, and automatable end to end. A web console is fine for setup. Production systems need a programmable surface.

The Giga CLI is that surface: for your engineers, your scripts, and your agents.

pip install giga-sdk

It’s available now for all Giga customers on PyPI. If you build something with it, or hit something that should be better, we want to hear about it.

GET A PERSONALIZED DEMO

Ready to see the Giga AI agent in action?

Giga's AI agents handle complex workflows at scale, from live delivery issues to compliance decisions, while maintaining over 90% resolution accuracy in production.

Introducing the Giga CLI — Giga