India AI Agent Funding Tracker 2026: Every Round, Sourced

Ask four Indian publications how much money AI startups raised in the first half of 2026 and you get four answers that are three times apart. Inc42 says $676 million. Entrackr says $2.07 billion. Tracxn says more than $1.42 billion for the year to September. None of them is wrong. They are counting different companies.

I wanted a narrower number: how much went to Indian startups that sell AI agents, the software that does a task end to end rather than answering a question. Nobody publishes that list for free, round by round. So I built it from the deal reports, one row at a time, and I will keep it weekly.

The rule is simple and it is written down in the methodology. A company counts if its main product is an AI agent, an agent platform, agent infrastructure or an agentic application, and if it is an Indian startup as the Indian tech press counts one. A round counts if the amount was disclosed and I could open a dated source for it. Foundation-model labs, GPU clouds, AI services firms and SaaS tools with an agent feature bolted on stay out, which is why the totals here are much smaller than the "Indian AI funding" headlines.

I build an AI agent platform in Bangalore, so this is also my own market. That is a reason to be careful with it, not a reason to flatter it. Where the record is thin, the page says so.

Every 2026 round

Newest first. Each company line shows its Indian base, and the headquarters too when that sits abroad. The amount is in US dollars as the source reported it, with the rupee figure under it. Click any source to read the deal report.

2026 to 4 October 2026: 45 rounds, $632.2 million. Rupee figures marked "reported" are the source's own; the rest are converted at ₹96.32 to the dollar (2 October 2026). "HQ" marks a company headquartered outside India. Download every row as data.csv.

Breakdowns

The same 45 rounds, cut four ways. All four tables are recomputed from the round table above on every pass, so they always agree with it.

By quarter

By city

Delhi NCR groups New Delhi, Gurugram and Noida. "No Indian city stated" covers rounds whose sources name no Indian city: mostly companies headquartered abroad that the Indian tech press counts as Indian startups.

By stage

Series A includes pre-Series A rounds and disclosed extensions; Series B and later includes pre-Series B. "Undisclosed stage" is a round whose amount was disclosed but whose stage was not.

By what the agents do

Why Indian AI funding numbers disagree

The three big Indian trackers each draw that line somewhere else. Entrackr's AI sector figure includes AI infrastructure and model companies, which is how one compute round of $1.2 billion can dominate a half-year. Inc42's AI sector is narrower and puts many AI-enabled products under their industry instead. Tracxn tags every company with AI in its description. Put the four side by side and the spread stops looking like a contradiction.

This tracker's H1 figure is recomputed from the table; the other three are as each publisher printed them (opened 4 October 2026). Tracxn's figure covers 2026 to its update date, not H1 alone.

If you are quoting a number about Indian AI funding, the useful habit is to quote the definition with it. "Indian AI agent startups raised X in H1 2026, counting only companies whose main product is an agent" is a sentence a reader can check. "Indian AI startups raised X" is three different sentences depending on who you read.

What the pattern says

The numbers in this section were written on 4 October 2026 and are not recomputed; the headline numbers and tables above are.

Most of the money goes to Indian startups with a US address

On 4 October 2026, 22 of the 45 rounds went to companies headquartered outside India that the Indian tech press counts as Indian startups, and those rounds held $502.2 million, 79 percent of the money. Emergent, Ema and Freehand all report a US headquarters alongside teams in Bengaluru or Chennai. The 22 rounds raised by companies headquartered in India add up to $123.0 million. Count only India-domiciled companies and more than half the rounds are seed or Series A; the largest is Equal AI's $30 million Series B.

Voice agents are the most crowded category

Eight of the 45 rounds were voice agents: Arrowhead, Bolna, Ringg AI twice, SimpleAI, Gnani.ai, Navana.ai and Dialflo. The cheques are small, $41.7 million across all eight, and the deal reports repeat one pitch: Indian languages, and phone-heavy industries such as banking, lending and insurance. Coding agents raised the most money, $221 million across six rounds, and Emergent's two rounds are $200 million of it.

Q2 was quiet, and Q3 made up for it

Q1 2026 had 21 rounds and $187.0 million; Q2 had 9 rounds and $87.5 million; Q3 had 15 rounds and $357.8 million. Part of the Q2 dip is real, and part may be coverage: the weekly tables I read for late May had almost no AI rows at all. If a curation pass finds Q2 rounds that I missed, they go into the table and the change log says so.

Methodology

What counts as a round

• Company. Main product is an AI agent, an agent platform, agent infrastructure (for example telephony built for voice agents), or an agentic application, in the deal report's own words or the company's. A CRM, analytics or SaaS product that adds agents counts only once the source describes the product itself as agentic.
• Indian. Headquartered in India, or headquartered abroad with Indian founders and a team, office or entity in India, as the Indian tech press reports it. The table says which.
• Round. Disclosed equity, pre-seed to late stage, announced in 2026. A disclosed extension or tranche is its own row and says so in the note (Ringg AI and Mykare). The date is the announcement date.
• Amount. USD as the source reported it. When a source printed a rupee figure, it is shown as "reported". Every other rupee figure, and every rupee total, uses one rate: ₹96.32 to the dollar, the European Central Bank reference rate on 2 October 2026.
• City. The Indian city the deal report names. When it names none, I use the company's Inc42 Datalabs profile and say so in the note; when neither names an Indian city, the row has none and the city table counts it separately.

Where the rows come from

The January to October 2026 seed was built from Inc42's weekly funding tables (every deal of the week, tagged by sector and sub-sector), cross-read against Inc42's individual deal reports and its Datalabs company profiles. Every AI and enterprise-tech row was read and classified by hand: 45 made it in. From here, a weekly news scan pulls candidate stories from Entrackr, YourStory, Inc42, the Economic Times and others, and each candidate is checked against its source before it becomes a row. Nothing is copied from paid databases.

Known gaps

• The seed leans on Inc42's coverage. A round that Inc42 never reported and that I have not yet found elsewhere is missing; the weekly pass is how that gets fixed.
• Very small rounds (under $500,000) appear only in weekly tables, without a standalone report, so their descriptions come from the company's Datalabs profile.
• Borderline cases left out on 4 October 2026, so you can see where the line sits: Smallest.ai (its deal report describes voice AI models), Hang Ten Systems (enterprise AI services), Rocketlane, Superleap, Bluecopa and GobbleCube (SaaS or analytics products adding agents), and Portkey (an LLM gateway). If one of them is described as an agent company in a later deal report, its later rounds will count.

How a correction gets logged

Rows are never edited silently. A changed amount, date or stage gets a dated line in the change log with the source that changed it.

The global tracker

This page is the India cut of a bigger record. The AI agent funding tracker for Q4 2026 follows every disclosed AI agent round worldwide, quarter by quarter, and the Q3 2026 edition holds the closed quarter. A round can appear in both: Emergent and Ema are on this page because of their Indian teams and on the global page because of their size. The two pages count differently, so do not add their totals together.

Two more trackers sit next to these: the AI agent pricing change log, which records what these companies charge and when that changes, and the AI agent incidents database, which records what goes wrong once agents are in production. For what agents cost to use today, the AI agent pricing index and the list of cheapest AI agent platforms are the places to start.

FAQ

Change log

1. 4 October 2026: Tracker opened with 45 verified rounds worth $632.2 million, dated 7 January 2026 to 25 September 2026, each read against its deal report before entry. Rupee conversions at ₹96.32 (2 October 2026).

Every weekly pass adds a dated line here. If a number on this page moves, the reason is in this list.

Sources