UK-based Portia AI raises €5.1M to help developers build safer, more reliable AI agents

London-based Portia AI, an open-source AI agents SDK (software development kit), has raised £4.4M (nearly €5.13M) in funding. The investment was led by General Catalyst, with participation from Firstminute Capital, Stem AI, and a group of angel investors.

The UK-based company emerged from stealth a few weeks ago and has since been working with early design partners to develop production agents and build out its SDK.

Building AI agents for production means

Founded by Emma Burrows and Mounir Mouawad, Portia AI has introduced an open-source SDK with a cloud component designed to help developers deploy AI agents in production. The company’s approach is built on three main pillars: predictability, controllability, and authentication.

Under the theme of predictability, Portia AI enables developers and users to define and review the intended behaviour of AI agents before any action is taken. This is made possible through its Planning agent, which allows plans to be expressed and adjusted in advance. The goal is to give teams visibility into how an agent will act, helping reduce the uncertainty often associated with deploying autonomous systems.

Controllability addresses the challenge of monitoring and guiding agent behaviour during execution. Portia’s Execution agents keep track of the plan’s current state and can pause to request human input through a system called Clarification. This interaction allows users to maintain oversight and influence over the process, which is especially important in environments where regulation, compliance, and transparency are required.

For authentication, Portia AI offers a cloud-hosted tool catalogue with built-in secure access controls. This allows users to authenticate agents within defined scopes, ensuring that agent activity remains aligned with organisational and user expectations.

Portia AI’s SDK is available to developers interested in building and deploying AI agents with visibility, oversight, and access control from the start.

Capital utilisation

Portia AI will use the funds to expand its developer team, speed up feature development, and enhance its SDK and cloud platform. The company aims to support easier and faster deployment of AI agents in production environments.

Portia AI’s Planning agent has been a central focus. The team recently introduced a new feature that allows users to “like” plans within the Portia dashboard. Based on those preferences, the agent retrieves relevant plans from Portia Cloud and uses them to guide its actions. This method enables agents to adapt plans from previous tasks to new user prompts. In internal tests, the Planning agent produced multi-step plans with full task coverage using only high-level instructions.

Portia AI is planning to release a step-by-step cookbook to help users explore this feature. The company is also exploring live A/B testing for multiple Planning agent variants, enabling users to match agent capabilities to specific needs. Early results from smaller local model deployments showed reduced tool selection errors by half.

Support for a wide range of language models has been added to the Portia SDK, including OpenAI, Anthropic, Mistral, Gemini, Azure OpenAI, and Bedrock. 

Users can also integrate their own local models, such as Llama or DeepSeek. In the coming months, Portia plans to support large input processing, including documents and complex API interactions.

Portia AI is also preparing to release a headless browser agent designed for secure user authentication. This feature will allow users to enter login details directly into browser sessions without sharing them with intermediaries. The agent will be able to resume tasks following human interaction during login.