We're building self-improving agents to automate tedious, repetitive operational work GREP.AI (YC F26). At IG, a UK online brokerage, Grep agents review compliance alerts, helping over 80% of traders onboard without manual intervention. IG is now a FTSE 100 company. Operational work often means following a specific procedure, gathering... information from different sources, checking it against what's needed, and deciding what to do next. The steps are repetitive, but no job is the same, making the work tedious to do by hand and hard to automate with workflows alone. Meanwhile, using frontier agents to reason through every task is often too expensive to scale. gives you an agent workforce that: -> Takes minutes to create and are domain experts with the right skills and tools from day one. -> Learn from every run, moving repeatable steps into code and workflows and saving frontier intelligence for exceptions and unfamiliar cases. -> Get smarter and cheaper as they scale using our AgentRun harness. -> Runs on our managed agent platform with secure sandboxes, audit logs, evals, cost controls, team sharing, and access controls included. Grep AI agents are available on demand and can run on schedules or in batch. Access them via the console, MCP, API, or SDKs that integrate into your existing workflows. Our customers, including IG, Airwallex, Worldpay, FV Bank, and Citizens Bank, have saved more than 250,000 hours of operational work with Grep Al agents across onboarding, doc verification, compliance, underwriting, operations, HR, GT ops, market research, and business intelligence. If operational bottlenecks are slowing your business down, create an agent workforce with Grep to automate them in minutes. P.S. We're in the Y Combinator Fall 2026 batch!show more

Learn more and build your agent workforce here:

@grepdotai Cool we open sourced a similar solution 4 months ago, and then Poetic raised 50M on the same primitive a few days later (agents into deterministic code): Works as a drop-in endpoint and built into all our agents. happy to chat!

@grepdotai This is interesting, I’m all for self improving agents. However, what stops OpenAI or Anthropic from building that self-improving loop into Codex or Claude Code themselves?

@grepdotai The 80% gets the headline, but I'd want to see the other 20%. When the agent hands off, does the reviewer get why it stopped and what it already checked? That handoff is where trust gets won or lost.

@grepdotai Yes we speed up those manual reviews too because all the research is already done for the analyst!

@grepdotai Congrats! Deeply useful product

@grepdotai moving the repeatable steps into code is the real trick. frontier on every task gets pricey fast

@grepdotai Compliance review is a great agent job because the output is already a decision plus a reason. The number I'd want next to the 80% isn't speed, it's how often the escalated 20% were the right ones to escalate. #AgentOps

@grepdotai How does an agent decide an alert needs frontier reasoning?

@grepdotai This shifts compliance from manual triage toward continuous monitoring and faster onboarding.

@grepdotai The part about agents learning from every run is pretty interesting. If they actually get better and cheaper over time, that could change how a lot of ops work gets done.

@grepdotai Congrats on the launch!

@grepdotai Congrats!!

@grepdotai log every agent decision, compliance teams ask for the trail

@grepdotai I'm sorry but this warped screen effect is so overdone. I wonder if opus could one-shot this 🤔 anyways. Congrats on the launch!

@grepdotai Haha it did but but def in the training data now!

@grepdotai Does each decision pin the harness version, policy version and source documents it used? With an agent that rewrites itself, reopening yesterday's case in today's harness isn't the same investigation.

@grepdotai Yes it does

@grepdotai Great launch 🚀, do you have benchmarks to show us?

@grepdotai Awesome product!!

@grepdotai congrats on the launch! curious, though, how do you decide what gets distilled into deterministic workflows versus what still needs frontier reasoning?

@grepdotai my bike app asks a model to identify a bike from its name. when the model was down or said not found, users got only the bike type back. now a web search runs as a second engine and a giant reign comes back complete in 6 seconds

@grepdotai looks great!

@grepdotai LFG!

@grepdotai 80% onboarding without a person means the remaining 20% are the hard cases by design. that's where a handover note from the agent pays off most.

@grepdotai The risk in moving learned steps into code is silent drift: the procedure or source format changes and the codified path keeps passing. Replaying a sample of codified runs through the full agent and diffing the decisions catches it early.

@grepdotai @andrewdsouza hey Andrew, take a look. Grep's agent workforce for ops onboarding and compliance. I can help them reach those banks.

@grepdotai Congrats on the launch, Grep is the future of operational work!

@grepdotai congrats on the launch!! 🚀

@grepdotai Can it do something like this?:

@grepdotai Yes we can automate this process. Reach out via DM

@grepdotai Great product!

@grepdotai compliance alert review is ideal agent work. Repetitive with a clear right answer and a human for the weird ones

@grepdotai Let’s goo!! Glad to make this video!