AI agents are attractive because they leverage LLM reasoning to offer autonomous decision-making while interacting with apps and data sources. However, many pilot projects haven’t reached production due to limited visibility into agents’ behaviour. Without the ability to monitor progress, intervene when needed, or securely authenticate agents into existing applications, teams lack sufficient control.
These limitations are particularly critical in regulated industries such as Financial Services and customer-facing applications where compliance, auditability, and trust are non-negotiable. Portia AI solves this problem.
Today, the company secured £4.4 million in funding to help developers build AI agents in production with better human oversight. The funding, led by General Catalyst, will accelerate the development of Portia’s open-source framework and cloud platform for building controllable, production-grade AI agents.
In what has been dubbed “the year of the agent,” Portia AI provides developers with tools to build production-grade AI agents they can predict, control, and authenticate. With this funding round, they’re expanding their developer team and preparing to release several exciting features throughout the year to create more reliable production agents.
Building frameworks for controlled and secure AI Agents
Portia AI was founded by Emma Burrows and Mounir Mouawad, two technology leaders with extensive experience in the fintech and tech industries. In March, the company emerged from stealth as a specialised AI platform that built frameworks for controlled and secure AI agents.
Mounir and Emma met at Stripe in London. As CTO of Stripe UK, Emma built the London engineering team from scratch and developed the e-commerce stack at Charlotte Tilbury. Mounir led several product expansions in EMEA, including bank-as-a-service at Stripe and the launch of Google Pay in 30 markets. Together, they bring high technical standards, stellar execution, and an unwavering focus on customer problems.
“AI engineers with deep expertise painstakingly navigate their way through the challenges of deploying reliable agents in production, but we want to democratise the solutions for all developers,” said Emma Burrows, Portia AI’s co-founder and CTO. Co-founder and CEO Mounir Mouawad adds, “Companies want to build products and processes on agents that feel like transparent, collaborative partners rather than black boxes.”
Portia AI tackles significant challenges in implementing AI agents for practical business functions. Clarifying and auditing agent actions minimises the likelihood of unforeseen behaviour. Human-in-the-loop validations guarantee that essential actions demand explicit approval or input instead of depending solely on the LLM. Detailed, real-time authentication ensures robust security while maintaining user-friendliness. Open-source resources and an accessible catalogue facilitate faster adoption and customisation.
The company’s name draws inspiration from literature. According to the founders, “Portia is our favourite protagonist from Adrian Tchaikovsky’s ‘Children of Time,’ which is also one of our mutual top three sci-fi novels!” The character Portia is part of an evolving spider species in the novel that advances its civilisation through technological and social innovation, representing resilience and connection—qualities the founders want to embody in their AI company.
What makes Portia AI special?
Portia AI stands out in the crowded AI agent landscape by focusing on three core pillars: predictability, control, and authentication. Its open-source framework helps developers and businesses build powerful and autonomous AI agents that are also transparent, steerable, and secure.
Portia AI enables agents to generate explicit, step-by-step plans responding to user prompts. Each step specifies which tools to use, what inputs are needed, and what outputs to expect. Execution agents carry out these plans while tracking the state at every step.
A unique feature of Portia AI is its “clarification” mechanism. Suppose an agent encounters missing information or ambiguous instructions or requires explicit approval (such as for a large financial transaction). In that case, execution can be paused, and structured human input can be requested before proceeding.
Portia AI provides a robust authentication layer, allowing agents to access external tools and APIS securely on behalf of users. This authentication is handled just-in-time, with token refresh, so agents only access what they need when needed—minimising security risks and improving user experience. The platform includes a growing catalogue of pre-integrated tools (e.g., Google, Slack, Zendesk, GitHub), accelerating deployment for common business workflows.
Developers can extend Portia’s capabilities by adding tools or customising agent behaviours. The open-source SDK and cloud dashboard make it accessible for various use cases, from fintech co-pilots to customer support automation. Every plan run is serialised and can be stored, retrieved, and audited.
“By seamlessly integrating AI agents, external tools, and human input – and by tackling challenges like JIT authorisation head-on – we believe Portia AI is paving the way for a new era of intelligent automation,” said Juliet Bailin, Partner at General Catalyst. “It also cannot be understated how brilliantly positioned Emma and Mounir are to build a generational company. Their vision is expansive, and they are well known among developers and business leaders alike for their ingenuity. We are exhilarated to be working alongside them.”
