LinearB launches Agentic Studio to aid coding agents

LinearB has launched Agentic Studio for engineering leaders and coding agents, adding a new context layer to its software delivery platform.

The launch comes as companies increase spending on AI coding tools while trying to speed the path from development to production. Agentic Studio is designed to turn data gathered across software delivery into information that both managers and automated coding systems can use.

Engineering leaders can query the system in plain language and receive answers and suggested next steps. Coding agents can access the same context, including information on past incidents, code ownership and rework history.

LinearB said its first "agent skill", now in early access, adjusts how much work a coding agent applies to each change. The system uses factors including code quality, recent service incidents and ownership history to decide whether a change is low risk or needs more careful testing and editing.

The company is pitching the product against a backdrop of rising AI-assisted software development and growing concern over whether more generated code is translating into more software releases. Research cited by LinearB, based on 2.7 million pull requests across 253 engineering organisations, found that developers using AI on at least 75% of coding days merged 96% more pull requests per week than a year earlier.

That increase has also created bottlenecks elsewhere in the development process. At the 75th percentile, AI-assisted pull requests were 2.6 times larger than unassisted ones, while pull requests opened by agents waited more than five times longer for review.

AI bottlenecks

The figures point to a familiar problem for software engineering teams adopting AI tools: code may be cheaper and quicker to produce, but review and release processes remain constrained by human oversight, team structure and ownership rules.

LinearB's data suggests autonomous agents still account for a relatively small share of changes. At the 90th percentile, agents opened 4.7% of pull requests. At elite organisations, 79% of agent pull requests merged within 30 days, compared with 92% of human-only pull requests.

At fair-performing organisations, the merge rate for agent pull requests fell to 37%, which LinearB attributed to weak ownership of AI-generated work. The company also said AI-assisted code review produced some of the strongest results, lifting merge rates by up to five percentage points.

The cost of AI use is also rising unevenly. Token spending at the 90th percentile reached USD $35.20 per developer per coding day, up nearly 75% in three months, while median spending changed little.

"Context is the new king," said Ori Keren, Chief Executive Officer and Co-Founder of LinearB.

"We realized we hold a knowledge graph that nobody else has of everything happening across the software delivery lifecycle, and that's exactly what agents and teams need to deliver high-quality software faster. So we rebuilt LinearB around it," Keren said.

Customer example

One customer, Kraken, used LinearB's MCP server with AI tools to examine delays in code review. According to LinearB, the analysis found that time zones and reviewer availability, rather than the size of code changes, were slowing approvals.

Kraken cut review time at the 90th percentile by more than half in four months, according to LinearB.

"That's the structural shift," said Nik Sudan, Engineering Operations Lead at Kraken.

"It moves leaders from anecdote-driven to evidence-driven," Sudan said.

The context layer draws data from Git systems and project management tools, links engineering identities across platforms and tracks changes from ticket to release. It also includes AI attribution and benchmarking data intended to help leaders assess whether internal performance is above or below that of peers.

Customers can use the context layer through LinearB's interface and through AI tools such as Claude via MCP. Agent skills work with any coding agent, the company said, and leaders can monitor how often agents are assigned low-, medium- or high-effort tasks.

Industry attention on this category has grown as engineering managers face pressure from boards and finance chiefs to justify spending on AI software. LinearB said one customer, Emburse, uses its data in board materials to answer questions about AI returns.

"These are exciting times to rethink our role in the software ecosystem," said Ori Keren, Chief Executive Officer and Co-Founder of LinearB.

"Everyone else is focused on writing code. We realized we hold a knowledge graph of everything happening across the software delivery lifecycle, and that's exactly what developers and agents need. So we rebuilt LinearB around it. We don't build the software factory. We instrument it so leaders know what to do next and agents make smarter decisions on every change," Keren said.