We shipped v2.0 less than two weeks ago, and shortly after we heard from a few early adopters: “I love Tako but can I just run this from my terminal?”
This feedback reflected a broader shift we’d been observing: CLI tools have become the standard interface for AI-powered developer tools. Claude Code, Gemini CLI, GitHub Copilot CLI—developers expect AI capabilities directly in their terminals where they’re already working.
Identity admins and SREs operate primarily from the command line. They needed Tako as part of their existing toolchain, not as a separate web application. And once they perfected complex queries, they needed those queries saved—not lost between sessions.
v2.1 delivers both: native CLI support and query memory.
What’s New in v2.1?
1. The Tako CLI
We built tako-cli.py to bring the full multi-agent orchestrator directly into your terminal. No browser, no UI, just pure command-line power.
Here’s what running a query looks like:
python scripts/tako-cli.py "List all users in the Engineering group with no MFA enrolled"
The agentic workflow runs—discovery, execution, and validation—all streaming live to your terminal. In most cases, you get results in under 5 seconds. When it’s done, you get:
- Results saved to CSV (in
logs/tako-cli-results/) ready to import anywhere - A preview of the first few rows printed to stdout
- The generated script archived in
logs/tako-cli-scripts/for debugging or re-running
It’s the same agentic power you get from the UI, but piped straight into your workflow. Chain it with grep, schedule it with cron, pipe it into other tools—it’s finally a native part of your shell.
2. Script-Only Mode (The Part I’m Most Excited About)
Sometimes you don’t want to run the query right now. You just want the Python code so you can review it, modify it, or integrate it into a larger automation.
That’s what --scriptonly is for:
python scripts/tako-cli.py --scriptonly "Generate a report of dormant admin accounts"
Tako will:
- Run the agentic workflow to discover the optimal query strategy
- Generate a fully executable Python script
- Save it to
logs/tako-cli-scripts/without running it
The --scriptonly flag gives you the code without execution—useful for reviewing the query logic, scheduling the script to run later, or integrating it into your existing automation workflows. The generated scripts can be used with other orchestration tools (Airflow, Jenkins, etc.) with some dependencies involved. Interested in portability for your use case? Reach out at support@fctr.io and we’ll help.
3. Query History & Favorites
The UI now tracks your last 10 queries. They show up in a collapsible sidebar, and you can re-run them with a single click.
But the real win is Favorites. If you’ve spent 15 minutes crafting the perfect query—something like “Show me all contractors with GitHub app access who haven’t logged in for 60 days and have no manager assigned”—you can star it. It stays pinned in your sidebar, ready to execute whenever you need it.
No more scrolling through Slack to find that one query you shared with your team three weeks ago. It’s just… there.
4. v2.0.5 Performance Boost (In Case You Missed It)
We shipped some major optimizations last week that are worth calling out:
- 3x faster sync operations: We tuned the concurrency limits to match Okta’s actual API tier limits (35 for Free tier, 75 for Enterprise). What used to take 10 minutes now takes 3.
- Smarter rate limiting: The agent now respects your specific Okta plan’s rate limit slider, so you’re not constantly hitting 429s.
- Data source transparency: The UI now clearly shows whether results came from the local SQLite cache or a live API call, plus when your data was last synced.
If you haven’t updated in a bit, this alone is worth pulling the latest version.
How Terminal Tools Fit the Bigger Picture
We’re seeing a massive trend right now where LLM-powered tools are moving out of chat UIs and directly into developer workflows. GitHub Copilot in your editor. Claude in your terminal. Cursor writing code where you actually work.
Tako started as a web tool because we needed a way to make complex Okta queries accessible to non-technical users. But the reality is that SREs, SecOps engineers, and identity admins already have terminal-first workflows. They don’t want to context-switch to a browser to check if a user has MFA enabled.
The CLI brings Tako into that world. You can now pipe Tako into bash scripts, schedule it with cron, or chain it with other command-line tools. It’s finally a first-class citizen in your existing toolchain.
What’s Next?
The architecture is in a really good place now. The Swarm handles complex queries efficiently, costs are under control, and the CLI means you can integrate Tako however you need.
Our focus for the next release is reliability and stability. We want to make sure the agent handles edge cases gracefully, that error messages are actually helpful, and that the system degrades gracefully when Okta’s API is slow or your local database is out of sync.
We’re also continuing to explore write operations (user modifications, group assignments, etc.), but we’re being extremely careful here. Expect strict approval workflows, dry-run previews, and human-in-the-loop controls before Tako touches anything beyond read-only queries.
Help Us Build This
Tako is built with the community, and it’s growing because people like you are using it, breaking it, and telling us what needs to be better.
If you’re using Tako or considering it for your team, star the repo on GitHub. It helps with visibility and lets us know this project is solving real problems.
Run into issues? Have ideas? Email us at support@fctr.io or open an issue on GitHub. We read everything, and your feedback directly shapes where this goes next.
Get Started
Ready to try it? The full setup still takes about 10 minutes. You’ll need:
- Docker (or Python 3.11+)
- A read-only Okta API token
- An LLM provider key (OpenAI, Anthropic, Google, or AWS Bedrock)
Full docs are on GitHub: https://github.com/fctr-id/okta-ai-agent
Thanks to everyone testing, breaking, and giving feedback on Tako. You’re shaping where this goes, and we’re building it with you.
—Dan & the Fctr team
IAMSE