Portia enables your developers to build safe, reliable AI agents in industries where compliance comes first. Create human readable plans for your agents, store where the agent is in its execution and define guardrails for when the agent should pause execution to authorise with a person or run additional checks. We support connections to hundreds of other products through pre-configured remote MCP servers and tool libraries, and also seamlessly handles the authentication for you.
Portia AI
Hey PH!
From our days dabbling with AI internally at Stripe, we started seeing the opportunities and challenges that regulated industries like fintech would face with AI. While reasoning and autonomy are powerful capabilities, LLMs are still stochastic in nature. How do you protect high stakes customer experiences where health, money and a ton of regulatory oversight are at stake? How do you enforce permissions and hand over to a human predictably? How do you keep evidence of all this to pass your audits?
We came out of stealth with our Python SDK in March and today we’d love to share with the PH community the full feature set we built up since then:
🧠 Transparent reasoning – Build a multi-agent `Plan` declaratively or iterate on one with our planning agent.
📝 Stateful execution – Get full explainability and auditability with the `PlanRunState`.
📍Compliant and permissioned – Implement guardrails through an `ExecutionHook` and raise a clarification for human authorisation and input.
🧰 1000s of cloud and MCP tools – Load any official MCP server into the SDK including the latest remote ones, or bring your own.
🛡️ Flexible deployment – Securely deploy on your infrastructure or use our cloud for full observability into your end users, tool calls, agent memory and more.
Want to give our SDK a spin? Head on over to our Github repository and give us a star while you’re there 💫
If your business is interested in building on our SDK, why not schedule a call with us ☎️
Thank you so much for giving us a try. Leave us a comment with your honest opinion so we can grow and learn.
–Mounir and Emma
Well Extract
@mounir_portia Go Mounir!
Portia AI
@maxime_champoux1 merci chef!!!
great to see a product focusing primarily on reliability and security.
while we adopt more AI agents into our daily workflow, this will play an important role.
congrats on the launch team!
Portia AI
@theanimeshs thank you so much for the kind words friend! let us know how you get on and shout on our Discord server if there's anything we can do to support you 🙏🏼
Portia AI
@theanimeshs Thanks for the support! keen to see what you'll be building!
Portia AI
@theanimeshs Thanks for your support! Love to hear about your AI agents and what use cases you have.
Congrats on the launch! Portia AI’s focus on compliance and transparent agent control is impressive. Could you share more about your typical user personas and the most common real-world use cases?
Portia AI
@techacc We're working fintechs primarily but are also seeing increasing interest from healthtech (and slowly but surely insuretech). A lot of use cases are human and natural language intensive back office functions that underpin UX e.g onboarding (KYC), ticket resolution (customer service), scheduling (after care).
@mounir_portia Does that mean you can support HIPAA workloads? Do you have a blanket BAA, or do you sign individual agreements?
Portia AI
Hey @techacc , appreciate your support! As Mounir mentioned, we've been working well with fintech and healthtech. However, the ambition is to serve a much wider range of use cases, especially in high-stakes, regulated industries where reliability and control matter.