Skip to main content
NeuronFeed
CATEGORY

AI Agents startups (2026)

Coding agents are clearing 70%+ on SWE-Bench Verified — the rest of the agent stack is still arguing about reliability

414 ai agents startups tracked, with the largest concentration in US. Total tracked funding: $107.9B.

Tracked
414
Total Raised
$107.9B
Countries
34
Active Deals
12

Editor's picks

6

Top by score

View all 414 →

Funding by year — AI Agents

2020 → 2026
$3.1M
’20
$52M
’21
$357M
’22
$3.2B
’23
$23.0B
’24
$45.2B
’25
$29.0B
’26

Market overview

The SWE-Bench Verified leaderboard turned over four times in 2025 alone. Anthropic's Claude Sonnet 4.5 pushed the public ceiling past 77%, OpenAI's GPT-5 agent traces matched it, and Cognition's Devin shipped solo runs that closed real GitHub issues at scale. Cursor ($7.6B raised, $29.3B valuation) and GitHub Copilot still own the IDE seat, but the open question for 132 tracked startups is whether anything outside coding ever reaches the same reliability bar.

The reliability ceiling

GAIA, WebArena, and OS-World show the gap. Coding clears 70%+ on Verified; general computer-use agents — Anthropic's Computer Use beta, OpenAI's Operator, Manus AI's autonomous browser — sit in the 40-60% band on multi-step web tasks. Decagon ($296M Series B) and Sierra ($1.27B Series C) sidestep the ceiling by narrowing scope: they run customer-facing dialogue inside a tight tool surface where every action is logged and reversible. Cognition's Devin chose the harder path and competes head-on with Cursor and Augment Code ($252M) on async engineering work where a wrong commit costs real money.

Where capital is concentrating

The top of the funding table is mostly cross-listed labs — xAI's $56B and Mistral AI's $6.3B carry over from Foundation Models because both ship agent surfaces. Among pure agent plays, Cursor sits ahead of Cognition ($1.14B Series C, $10.2B valuation), Glean ($750M Series F), Notion AI ($343M Series C), and Imbue ($232M Series B). Andreessen Horowitz, Sequoia, Thrive, and General Catalyst lead most rounds. Cline raised $4M seed and ships a VS Code agent that competes with $7B-funded incumbents — a useful proof that distribution and BYO-model flexibility still beat raw capital in this segment.

What 2026 actually decides

Whether agent reliability transfers. Coding works because compilers fail loudly, tests are cheap, and rollback is git reset. Booking a flight, filing an expense, or modifying a Salesforce record has none of those properties. The startups building eval harnesses, recovery loops, and explicit human-handoff UX — not the ones shipping flashier browser demos — are the ones likely to be alive in 18 months.

Key trends 2026

  • SWE-Bench Verified is the only benchmark anyone trusts. Claude Sonnet 4.5 and GPT-5 agent traces both cleared 77%, but only on coding — GAIA and OS-World still sit 30 points lower for general computer-use.
  • Anthropic Computer Use, Operator, and Manus AI compete on the same OS-World task set. Public scores hover 40-60%; nobody has shipped a horizontal agent at coding-grade reliability yet.
  • Vertical narrowing beats horizontal autonomy commercially. Decagon ($296M) and Sierra ($1.27B) win deals by constraining tool surface; Devin and Augment Code win by going deep on engineering specifically.
  • Open-source agents punch above weight. Cline raised $4M seed, runs in VS Code, and ships against products with 1000x the capital — distribution still routes around funding.

Benchmarks vs global

SWE-Bench Verified ceiling
77%+
Claude Sonnet 4.5 / GPT-5 agent traces
OS-World general computer-use
40-60%
Operator, Computer Use, Manus AI band
Total funding tracked
$87.9B
second-largest category by capital
Companies tracked
132
43 US HQs (33%), 4 China

Top countries

By startup count

Stage breakdown

Latest round type
  • Seed 182
  • Series A 65
  • Series B 33
  • Pre-Seed 23
  • Series C 21
  • Series D 10
  • Series A Extension 3
  • Corporate 3

Top investors backing AI Agents

See all →

FAQ

Frequently asked

Which AI agents actually pass SWE-Bench Verified above 70%?
As of late 2025, Anthropic's Claude Sonnet 4.5 and OpenAI's GPT-5 both posted public agent traces above 77% on SWE-Bench Verified. Cognition's Devin reports comparable numbers on its full agent harness, and Cursor's Composer mode clears 60%+ in independent runs. Outside coding, GAIA and OS-World numbers for Operator, Computer Use, and Manus AI still sit in the 40-60% range — the reliability gap between code and general computer-use is roughly 20 points.
What separates AI Agents from AI Chatbots in this directory?
Chatbots reply; agents commit side effects. A chatbot drafts the email; an agent sends it, files the response in CRM, and books the follow-up. The same Claude or GPT model can power both — the difference is the scaffolding around it: tool registries, browser control, persistent memory, and recovery logic. NeuronFeed lists 132 agent startups separately from 50 chatbot startups because the buyer expectations and reliability bars are not comparable.
Why do investors keep funding coding agents specifically?
Coding is the one domain where agent failure is cheap and detectable. Tests run, compilers throw, and `git revert` exists. Cursor at $29.3B, Cognition at $10.2B, Augment Code at $977M, and Imbue at $1B reflect that revenue actually shows up — engineering teams will pay $20-40 per seat for a Copilot that ships PRs. No other agent vertical has a comparable feedback loop yet.
Is Manus AI competitive with Operator and Computer Use?
On public OS-World numbers, all three sit in roughly the same band. Manus AI scaled fastest on consumer demand in early 2025; OpenAI's Operator has the GPT-5 reasoning trace; Anthropic's Computer Use is the API path enterprises actually deploy. None of them are reliable enough yet for unattended 50-step workflows on production data — that gap is the central commercial bet of the category.
Where are agent startups concentrated geographically?
The US holds 43 of 132 published HQs (33%). China contributes 4 entries including Zhipu AI and Moonshot AI on the foundation side. Israel has 3, with Sweden, the Netherlands, and Czechia (JetBrains) at 1 each. The pattern mirrors AI Developer Tools — most of the agentic frontier is currently coding-focused, and that work tracks where the model labs and IDE incumbents already operate.

Recent rounds in AI Agents

All rounds →
Date Startup Round Amount
Jun 2026 Lassie Series A $35M
May 2026 Cognition Series D $1B
May 2026 Flick Seed $6M
May 2026 Outmarket AI Series A $17M
May 2026 Vapi Series B $50M
May 2026 Exaforce Series B $125M
May 2026 Moonshot AI Series C $2B
May 2026 Fazeshift Series A $17M

All AI Agents startups

Page 1

Resolve AI

US est. 2023

AI SRE for complex production environments

Raised
$190M
Stage
SERIES A EXTENSION
90

deepset

DE est. 2018

Haystack framework and deepset Cloud for enterprise LLM apps

Raised
$45.6M
Stage
S-B
82

Wordware

US est. 2023

Build AI agents by writing in plain English

Raised
$30M
Stage
Seed
81

Wonderful

IL est. 2025

Multilingual enterprise AI agents for customer service

Raised
$286M
Stage
S-B
80

Cursor

US est. 2022

The AI code editor built for productive engineers.

Raised
$7.6B
Stage
S-D
79

Oasis Security

IL est. 2022

Agentic access management to secure AI agents and non-human identities

Raised
$195M
Stage
S-B
78

Emergent

IN est. 2024

Agentic vibe-coding platform that builds full apps from natural language

Raised
$30M
Stage
S-A
74

Alaan

AE est. 2022

AI-powered spend management and corporate cards for MENA finance teams

Raised
$48M
Stage
S-A
74

Enter

BR est. 2023

AI to manage corporate litigation in Brazil's complex legal system

Raised
$35M
Stage
S-A
74

Serval

US est. 2024

Dual AI agents that automate enterprise IT service management safely

Raised
$127M
Stage
S-A
74

Contextual AI

US est. 2023

Build specialized RAG agents for the enterprise

Raised
$100M
Stage
Seed
74

Luminance

GB est. 2015

Legal-Grade AI agents that automate contract drafting, negotiation and review.

Raised
$165M
Stage
S-C
73

Model ML

US est. 2024

AI workspace that automates investment banking and finance workflows

Raised
$87M
Stage
Seed
73

Exaforce

US est. 2023

Agentic SOC platform for real-time threat detection and response

Raised
$200M
Stage
S-B
72

Liberate

US est. 2022

Reasoning AI agents that automate insurance sales, service and claims

Raised
$72M
Stage
S-B
72

ZyG

IL est. 2024

An agentic operating system that scales DTC ecommerce brands end to end

Raised
$118M
Stage
S-A
72

Augment

US est. 2024

AI teammate that automates freight order-to-cash workflows

Raised
$110M
Stage
Seed
72

Kognitos

US est. 2020

Neurosymbolic agentic automation that runs business operations in plain English

Raised
$51.2M
Stage
S-A
72

Poke

US est. 2024

An AI personal assistant that lives inside your text messages

Raised
$25M
Stage
Seed
72

Crescendo

US est. 2024

AI-native contact center blending agents and humans

Raised
$50M
Stage
S-C
71

Mastra

US est. 2024

The open-source TypeScript framework for building AI agents

Raised
$13M
Stage
Seed
71

Lio

DE est. 2022

Multi-agent AI system automating enterprise procurement end to end

Raised
$30M
Stage
S-A
71

Composio

US est. 2023

Integration infrastructure that lets AI agents take action across 250+ apps

Raised
$29M
Stage
Seed
71

OpenHands

US est. 2024

Open-source AI software agents that plan, write, and ship code across entire codebases.

Raised
$18.8M
Stage
S-A
70