Agentic Index

Best coding agents ranked on the merge loop (2026)

Data last verified October 21, 2026

The Agentic Index grades 78 coding agent vendors against the same 14 capabilities as every other lane. These are independent ratings of AI coding agents, rated from each vendor's own public evidence rather than from analyst opinion, vendor briefings or paid placement. 2 of them document the full merge loop. That means the agent reads the codebase, reaches the systems the work actually lives in, verifies its own change before a human sees it, leaves a record somebody can read afterwards, and a human governs what reaches the default branch. That is 2.6% of the lane. This is a bar, not a leaderboard: a vendor either documents all five in its own public materials or it does not clear, and Partial evidence on any one does not count.

Before the list, the thing worth knowing about the list. Only one of the 2 vendors that clear this bar writes the code. Baz is code review, the layer that checks the work. Zencoder is the one code writing platform in the lane that documents the whole loop, and it gets there by pairing its building agents with a dedicated testing agent, Zentester. Every other vendor whose product writes code fails, including the highest coverage vendors in the pool: GitHub Copilot 13.5, OpenHands 13.5, Base44 13.0, Cognition 13.0, Kilo Code 13.0, Kiro 13.0, every one of them at or above the 13.0 carried by the vendors that clear.

The finding underneath that is the reason the third capability is on the bar. This field can write the code and cannot prove it works. Testing, debugging and optimization sits at 22 of 78, the thinnest axis on the bar by a distance, and it is the sole blocker for 9 of the 16 vendors sitting one capability short. 24 vendors document both reading the codebase and reaching your systems in full, and 16 of those 24 do not document self verification. That is the cleanest single test to run in a demo: ask to watch the agent run the suite, read a failure and fix its own work with nobody prompting it.

Two more numbers to carry into a vendor call. The field asks permission and then does not check its work. 44 of 78 document human oversight and guardrails in full, at least twice the 22 documenting self verification, and 34 of those 44 document no testing. That ordering is backwards for the job. An approval gate in front of an unverified change does not reduce review burden, it moves the whole verification task onto the reviewer, who is the person the agent was bought to spare. And memory is thinner than the category suggests. 19 of 78 document memory and state persistence in full, about one in four, and neither of the 2 that clear is among them. For a category built on repository context that is worth asking about. Most of what reads as memory in this lane is a sandbox that keeps its files or a rules file the product applies, rather than a store the agent maintains with a stated scope and lifetime.

One honest note on the bar itself. 2.6% is harsh, and it is deliberate. Drop self verification and the list gets much longer and much less useful, because a coding agent that writes a change and cannot document checking it has handed a reviewer the exact part of the job they were delegating. The bar stays at five for that reason. Disagree with it if you like: every count on this page is stated so you can recompute the result against four.

The merge loop, and how the 78 vendors score against it

Step	Capability, and what has to be documented	Full	Only blocker
1.It reads the codebase	Knowledge grounding and RAGThe agent grounds itself in the actual repository, its conventions and its history, rather than reasoning over the file currently open. Without this it is guessing in your dialect.	28 (36%)	5
2.It reaches the systems	Integrations and tool callingIt calls into the repository, the build, the test runner, the issue tracker and CI, so the work lands where the work lives rather than in a chat window.	63 (81%)	1
3.It verifies its own change	Testing, debugging and optimizationIt runs tests, reads the failure, retries and gates its own output before a human sees it. This is the capability that separates an agent from an autocomplete, and it is the thinnest on this bar.	22 (28%)	9
4.It leaves a record	Observability and auditabilityA durable trace of what the agent read, what it changed and why, readable by the person reviewing the diff and by whoever inherits the code a year later.	37 (47%)	1
5.A human governs what lands	Human oversight and guardrailsAn approval step or scoped permissions the vendor ships, on anything that reaches the default branch. The customer's own pull request merge or branch protection is the customer's control, not the vendor's, and does not count on its own.	44 (56%)	0

Full means the vendor publishes evidence meeting the capability in its own public materials, under the Agentic Index verification standard. Only blocker counts vendors that document the other four capabilities in full and fail on this one alone. Read the last two columns in order: tool calling reaches 63 of 78, grounding 28, oversight 44, then auditability drops to 37 and self verification to 22, where it accounts for 9 of the 16 near misses on its own. The field can read your repository and change it. What it cannot consistently document is checking the change.

The 2 vendors that document the full merge loop

Ordered by total coverage across all 14 capabilities, ties broken alphabetically, which is exactly what the rankings pages compute. Coverage measures breadth across the whole taxonomy, so a lower score here usually means a narrower product rather than a weaker one. Every entry says what it actually operates on, beside its name, because clearing this bar does not mean a vendor writes the code.

1. 1.Baz

   13.0 / 14 capabilities

   Operates on code by: reviews and governs code before and after it is written, including code written by other agents

   Level with Zencoder at the top of this list, and the cleared entry that reviews code rather than writing it. Baz runs agents directly on the codebase to review, govern and secure what developers and AI copilots produce: a code review agent on every pull request that detects breaking changes across endpoints and behavior, a Spec Reviewer that runs the app in a sandbox and drives a browser to check the built experience against Figma and Jira, and Planner, which reviews and approves the plan before any code is written. The founding team came out of Palo Alto Networks by way of Bridgecrew, and it extended a seed round to $17 million in June 2026 co led by Battery Ventures and Boldstart. Its stated premise is the one this page keeps arriving at from the data: when AI generates code faster than a team can read it, reviewing after the fact stops being a defense. Pricing is a free trial and an on demand starter on the GitHub Marketplace, with team and enterprise tiers not fully published.

2. 2.Zencoder

   13.0 / 14 capabilities

   Operates on code by: writes code with building agents and checks it with a dedicated testing agent before deployment

   Repo Grokking indexes code across the whole organization and writes a persistent repo.md, build, review and audit agents each run on a different model, and Zentester generates unit and end to end tests in the project's own frameworks. The agent pauses for approval on actions unless a user switches to auto permission mode, so the gate is on by default and can be turned off.

The 16 vendors that miss by exactly one capability

These document four of the five and fail one. In this lane it is also the list of every name you have heard of, and GitHub Copilot sits at the top of it carrying the joint highest coverage score in the entire pool. If you are already holding a shortlist this table is the most useful thing on the page, because it tells you the single question to ask each vendor rather than sending you back to the start. A miss is a documentation finding, not a verdict on the product: several of these vendors may well do the thing and have not published evidence that meets the standard.

Common questions

What are the best coding agents in 2026?

Two of the 78 vendors in the Agentic Index coding agent lane document the complete merge loop in full: the agent reads the codebase, reaches the systems the work lives in, verifies its own change, leaves a readable record, and a human governs what lands through a control the vendor ships. Ordered by total coverage across all 14 capabilities they are Baz and Zencoder. That is 2.6% of the lane. Baz reviews code rather than writing it. Zencoder writes it, and pairs its building agents with a dedicated testing agent. Sixteen more vendors sit exactly one capability short, and that table is where most of the names you have heard of appear.

Why does Claude Code, Cursor or GitHub Copilot not appear on the list?

Each falls short on the public record, and not on the same capability. GitHub Copilot documents four of the five and fails one, testing, debugging and optimization, while carrying the joint highest total coverage in the pool at 13.5 of 14, above every vendor that clears. Anthropic Claude Code documents four of the five and falls short on reading the codebase, where its grounding evidence is partial. Cursor falls short on both of those, as do OpenAI Codex, Google Jules and Poolside. That is not a claim that these tools cannot run a test or read a repository. It is a finding about what each vendor publishes, and the question to put to the vendor is narrow and answerable.

Why is self verification on the bar at all?

Because it is the capability that separates an agent from an autocomplete. A coding agent that writes a change and cannot show that it checked the change has handed the reviewer the exact part of the job the reviewer was trying to delegate. It is also the thinnest axis in the lane by a distance, documented in full by 22 of 78, and the sole blocker for 9 of the 16 vendors one capability short. Running the same pool against a four capability bar without it produces a much longer and much less useful list, which is why the bar stays at five and why this answer exists.

How many coding agents can actually act on a repository?

Of the 78 vendors, 24 document both knowledge grounding and integrations and tool calling in full, meaning they read the real codebase and reach the repository, build, tests and tracker. Of those 24, only 8 also document self verification. So 16 vendors can make a change in your systems and do not document checking it first. That is the single cleanest question to take into a demo: ask to watch the agent run the test suite, read a failure and fix its own work without anybody prompting it.

Do these tools have human approval before code is merged?

Of the 78 vendors, 44 document human oversight and guardrails in full, at least twice the 22 documenting self verification. That ordering is backwards for the job. An approval gate in front of an unverified change does not reduce review burden, it relocates the whole verification task onto the reviewer. 34 of the 44 vendors that document oversight do not document testing, and those 34 are where the review load is quietly landing.

How is this different from a coding agent benchmark or leaderboard?

A benchmark measures how well a model completes tasks. This measures what a vendor documents about how its agent operates inside your engineering process, graded against a fixed 14 capability taxonomy from the vendor's own public materials, with no vendor briefings, no submissions, no paid placement and no vendor review before publication. The bar is mechanical and stated in full so you can disagree with it and recompute it. It also names what each vendor fails rather than only what it does, which is why the near miss table is eight times longer than the list above it.

Method. This page runs the horizontal grading method over one lane of the index: membership is the Coding agent category, 78 vendors out of 1,078 researched, 63 by primary category and 15 by secondary, every one carrying a complete 14 row capability grid, with none excluded for incomplete evidence. The bar is five of those 14 capabilities at Full, chosen because they are the loop a change actually travels, and it is stated in full above so you can disagree with it and recompute. Every grade comes from the vendor's own public materials under the Agentic Index verification standard. No vendor pays for placement and no vendor has reviewed this page. Data last verified October 21, 2026. How this evidence is graded

Related: all 78 coding agents ranked on total coverage, the full coding agent capability matrix, how every vendor scores on testing and self verification, how every vendor scores on knowledge grounding, Cognition against Factory, Charm against OpenCode, CodeRabbit against Graphite, Cognition against Cursor, agent observability platforms, where the record gets read, platforms that require human approval before an agent acts, production ops agents ranked for automated incident resolution, agentic security operations platforms ranked for alert triage, compare vendors side by side.