The new startup wants to help developers create customized context coding assistants that can connect with any model and integrate seamlessly with the development environment.
Founded in June 2023 by CEO Ty Dunn and CTO Nate Sesti (pictured above), Y Combinator Alum Continue has gained approximately 23,000 stars and 11,000 Discord community members on Github over the past few years. To build on this momentum, Continue has announced version 1.0 of its product, supported by a new seed funding of $3 million.
Coding assistant explosion
The launch of Continue comes amid the explosion of AI coding assistants such as Github Copilot and Google's Gemini Code Assist. This goes without saying that young startups like Cordum and Cursor have raised bucket loads from investors.
Continuingly, some of them will pitch themselves as “primary open source AI code assistants” that can connect to any model, allowing teams to add their own context by pulling data from platforms like Jira and Confluence.
The model and context are connected, allowing developers to create custom autocomplete and chat experiences directly within their coding environment. For example, AutoComplete offers inline code suggestions when entering, but chat allows users to ask questions about a particular code. The editing feature also allows users to modify the code by explaining which changes they want to make.
Continue with ActionImage Credit AI Coding Assistant: Continue
The product facets announced today include the first “major” release of Continue's open source extension for VS Code and Jet Brain.
“This shows companies that this is a stable project you can wager and build,” Dan told TechCrunch in an interview.
Separately, Continue launches a new hub. This can be compared to things like Docker Hub, Github, and Hugging Face. This is where developers create and share custom AI code assistants and replenish the registry to define and manage the various building blocks they have created.
At launch, the hub includes a pre-built AI coding assistant, as well as a validated partner Mistral and its codestral model, Claude 3.5 Sonnet, and “blocks” of Ollama's Deepseek-R1. However, individual vendors or developers can donate blocks and assistants to the hub.
The block here could mean a model that allows you to specify the AI model and location to use. Rules for customizing your AI assistant. Contexts that define an external context provider (e.g. Jira or Confluence); prompts to pack pre-built models use prompts to invoke complex instructions. Documents that define a document site (Angular or React); data that allows developers to send development data to predefined destinations for analysis purposes. Or an MCP server that defines standard ways to build and share tools for language models.
Continue your hubimage credits: Continue
“Culture of Contribution”
The idea behind this new hub is that the majority of users do not need deep customization. This means that you need to make minor tweaks to the coding assistant or blocks that already exist in the hub.
This raises the question: what are the incentives for creating customizations and sharing them with the world? After all, it really drives the open source community elsewhere. Many of the launch partners are the very companies that create the underlying tools and models (such as Mistral and Anthropic), and Continue's New Hub makes it an ideal place to curry favors with developers.
Furthermore, the “open source spirit” is at the heart of the ongoing thing. If someone creates customizations for use in the workplace, why don't they share them with the wider community? Ultimately, Continue positions itself as the antithesis of its own “black box” AI assistant provider.
“This is a hub for the entire ecosystem to come together and work together,” Dan said. “What if instead of everyone building their own closed source AI code assistants, we all worked together to create the building blocks that people need to build tailored experiences for themselves?”
This is what Dan calls an establishment of a “culture of contribution” and encourages developers to experiment and create their own customizations while creating value for everyone.
“Continuing 1.0 allows this contribution culture that allows developers to create and share custom AI code assistants,” says Dunn. “This registry will become a place of discovery within and across organizations. It will grow with developer tools and locksteps enhanced with block evolution and open AI.”
Next is the data control aspect. On a more common “all-purpose” platform, vendors can extract great value from observing how developers work at scale, bring this decision data back to the platform and improve things for everyone. This type of activity has sparked controversy over things like Github Copilot. This has been accused of hijacking the efforts of millions of open source software developers for their own benefit.
The idea that if followed, companies have more control over what happens with data.
“If you continue, you can keep your data,” Dan said. “As an organization, we can pool all data from all developers in one place. This is not possible with a black box code assistant of all sizes. Here, the SaaS offering and strategy is to get the data and use it to improve for everyone.”
Model Business
It's still relatively early to continue, but the startup says it has worked through a few well-known businesses, Ionos (early ongoing customers), and Siemens and Morningstar throughout the development phase.
Large companies are very focused, but Dan says he targets developers of all shapes and sizes, from freelancers and small teams to companies. This shows how continuity makes money – its new hub features a free solo tier, but organizations that need more control over their data can pay to access additional management, governance and security tools.
“There's a lot of interest from large organizations, but I've seen it all with individual developers who want some sort of customization for themselves. In those cases, I think Solotia is more than enough,” Dan said. “But that freelancer and small teams start to grow and require some degree of governance, so you can become a client.”
Free Solotia ships with three “visibility” levels. Developer contributions can be shared internally as part of the team or fully publicized. In fact, solo tiers can be technically used in team setups. There is a lack of some of the features that teams normally require. Another “Team” tier adds additional “Multiplayer” smarts to the mix, using admin controls to manage all blocks and assistants.
Meanwhile, Enterprise Tier will increase data, security and governance options with more detailed control than those using blocks, models, versions, and vendors.
“Admins can manage security regarding the credentials that the data progresses and receive audit logs from developer usage, what, when, and where,” says Dunn.
The rest was a $2.1 million vault (finance with delayed share allocation) that had previously raised $2.1 million in late 2023, but was led by Deavebit, a developer-centric VC company.
Dan says the majority of the fresh cash will be directed towards software engineering salaries, with the current five personnel planning to “at least double the number.”
“We use open source as our distribution approach, so we keep costs very low. We don't need to capitalize almost as much as other competitors,” Dan said.