1st place
$2,000cash
+ $5,000Dodo Payments credits
#neatHackAI agent hackathon
Build agents you’d trust to run without you watching.
Powered by
Hackathon starts in
- Dates
- October 10–11, 2026
- Track
- One track
- Format
- Online · 48 hours
- Prizes
- $19,000 USD including credits
Partners
Get started
Register for neatHack
Join the neatHack Discord
Check in and confirm your team
Submit your project
Monday, October 12 · 7:00–11:55 PM IST
Submissions opening soon
Prizes & people
$4,000 in cash, plus partner credits and neatlogs Pro.
2nd place
$1,000cash
+ $3,000Dodo Payments credits
3rd place
$500cash
+ $1,000Dodo Payments credits
Each winner gets 3 months of neatlogs Pro (worth $1500).
Community winner
$500cash
A separate award for community impact. You can win this alongside a top-three placement.
Credits to build with
- AWS creditsFirst 200 teams to complete web check-in
- 5,000 creditsEveryone participating
The only track
Build an agent that works (literally)
Build an AI agent, or a team of agents, that completes a substantial task from start to finish. Give it a goal, tools and an environment; let each result guide its next action. All projects compete for the same prizes.
Focus areas
Planning
Break a goal into actionable steps.
Tool use
Choose tools, supply inputs and handle responses.
State
Retain state across steps, runs and failures.
Context
Bring in relevant information; leave out the rest.
Execution
Keep long tasks running to completion.
Recovery
Handle failures and unexpected results.
Verification
Check that results are correct.
Iteration
Use each result to choose the next step.
See example tasks
These examples show the size of task we have in mind. They aren't separate tracks, and you don't have to build one of them.
- Fix a GitHub issueAn agent investigates a GitHub issue, writes and tests a fix, and proposes the fix.
- Answer a research question with sourcesAn agent splits a question into smaller questions and gathers evidence for each one. When sources disagree, the agent works out which source to trust and writes an answer with citations.
- Watch a running systemAn agent watches a running system for unusual behavior. When the agent finds a problem, it looks for the cause and takes an action to fix it.
- Several agents working togetherSeveral worker agents each solve part of a task on their own. A coordinator agent then combines their results and settles any disagreements.
- Work while things changeAn agent works on a long task while the environment it works in keeps changing.
Requirements
neatlogsWorks with any agent
Trace every agent run
Find the code behind a failure, then compare runs to verify your fix.
- Model and tool calls, inputs and responses
- State changes, retries, failures and agent messages
- Latency, token usage and cost
Research agentEvery step, in context.
SuccessSpan treeDuration
- agentresearch_agent2.84s
- llmplan620ms
- toolsearch_documents480ms
- llmcompose_answer1.42s
- toolverify_citations320ms
EntireFor coding agentsGive your coding agent repository context
Open-source tools for understanding your codebase and the history of AI-made changes.
- Checkpoints: save the context behind each change
- entire-graph: map relationships across your repository
Preserve state across retries #42
agent/recovery → main
Conversation 2Files changed 1+3 −1
src/agent/run.ts
− const state = createState();+ const state = checkpoint.restore();+ await agent.run(state);+ checkpoint.save(state);Plan the business with cfo.ai
Use cfo.ai to plan your agent’s pricing, costs, revenue and runway. Include the resulting business plan in your submission.
C6ƒx
=(C3-C4)/C3| A | B | C | |
|---|---|---|---|
| 1 | Monthly model | Base | Growth |
| 2 | Customers | 100 | 250 |
| 3 | Revenue | $2,000 | $5,000 |
| 4 | Infra + inference | $400 | $850 |
| 5 | Gross profit | $1,600 | $4,150 |
| 6 | Gross margin | 80% | 83% |
Judging criteria
Judges score each project in five categories. Each category is scored out of 100, and the categories are weighted as shown to give a final score out of 100.
The agent
How well the agent, or the system of agents, handles a substantial task from start to finish.
Use of neatlogs, Entire and cfo.ai
How effectively you used neatlogs, Entire and cfo.ai to build, improve and plan a business around the agent.
Demo video
How clearly the demo video explains the agent and your use of neatlogs, Entire and cfo.ai.
Building in public
How you shared your project with other people while you built it.
Usefulness and originality
Whether the agent solves a problem that people have, in a new way.
Evidence & verification
Back up each claim in your submission with evidence that judges can check.
- The agent improved
- A run before your change and a run after it, e.g., two neatlogs traces, with the numbers you compared
- You used neatlogs
- Links to neatlogs traces from your project, or screenshots of them
- You used Entire
- Checkpoints or entire-graph output from your repository
- You used cfo.ai
- A business plan covering your agent’s pricing, costs, revenue and runway
- The agent recovers from failures
- A run where an action failed, showing what the agent did next
- You built in public
- Links to what you shared during the hackathon, tagged #neatHack
- You wrote the code during the hackathon
- A public repository with commits dated during the hackathon
Schedule
Build October 10–11, submit October 12. Winners are announced October 16. All times below are in IST.
Day 1 · kickoff
- Intro call on Discord
9:00 AM IST — Meet the community and get ready to build.
- Kickoff
12:30 PM IST — Welcome, the challenge, partner requirement, rules, judging and sponsor introductions.
- Workshops & office hours
Technical sessions and support from neatlogs and sponsors.
Workshops & submission
- Final workshops & office hours
Sunday, October 11 — The last round of technical support before submissions.
- Submit & share on X
Monday, 12 October 2026 · 7:00–11:55 PM IST
Submit on X by quoting the official neatlogs post with your team, name, project, public repo and demo. neatlogs and cfo.ai will engage with the first 50 quote tweets.
Results
- Winners announced
We announce the three placed winners and the community winner.
Judges & mentors
Judges
Mentors
Building resources
Optional services for model access, infrastructure and data.
Access AI models through Amazon Bedrock. Use AWS compute and storage to run your agent and keep its data.
AWS credits for the first 200 teams to complete web check-in
Give your agent company and people data through Crustdata’s APIs. Useful for research, discovery and enrichment tasks.
5,000 credits for everyone participating
Questions & rules
FAQs01 / 10
When and where is neatHack?
neatHack is a 48-hour online hackathon on October 10–11, 2026, starting Saturday at 12:30 PM IST. The intro call is at 9:00 AM IST on October 10. Submissions open October 12 at 7:00 PM IST and close at 11:55 PM IST. Winners are announced on October 16.
FAQs
neatHack is a 48-hour online hackathon on October 10–11, 2026, starting Saturday at 12:30 PM IST. The intro call is at 9:00 AM IST on October 10. Submissions open October 12 at 7:00 PM IST and close at 11:55 PM IST. Winners are announced on October 16.