NEAR
KLEAR3 ResearchPublished Updated
NEARNEARNEAR is a sharded, proof-of-stake blockchain that runs its own smart contracts rather than a token riding on someone else's chain. Its repository has published code since October 2018, and GitHub's own licence badge for it is wrong: it says GPL-3.0, a licence with no matching file anywhere in the project.
What does NEAR actually do?
NEAR runs smart contracts on a chain split into shards, so more transactions can be processed in parallel instead of one at a time. Most of every gas fee is destroyed rather than paid to a validator, and the remaining share goes to the account that executed the call.
Sharding is the part worth spelling out, because it is the actual engineering bet behind the throughput number, not a marketing word. The chain is cut into pieces that each validator set can process, and a design the project calls Nightshade is meant to let that split grow without forcing every machine to hold the whole ledger at once.
That is a different bet from the chain that made programmable ledgers ordinary, which instead pushes heavy computation onto separate rollups and keeps its own base layer simple. Neither approach is settled science, and this page does not pretend one has won.
Source near.org, read 2026-10-05
Is the code actually open?
Published, yes, and its own licence badge is wrong. GitHub reports GPL-3.0 for the repository. No file matching that licence exists anywhere in it: the project's own build file, and every licence file actually shipped inside the code, declare MIT or Apache-2.0 instead.
This is not a missing file, which is what Pendle's repository shows. It is the opposite trap: a badge that confidently names the wrong licence, on a repository that answers the question correctly itself, repeatedly, if anyone opens the files GitHub is supposedly reading.
The project's own Cargo.toml states license = "MIT OR Apache-2.0". Walking the full repository tree, over 50 separate crates each carry their own LICENSE-MIT and LICENSE-APACHE files. Not one of the roughly 3000 paths in that tree contains GPL text. A reader who trusts the badge and avoids the project for copyleft reasons is avoiding it for a reason that does not exist.
Source GitHub API vs. Cargo.toml et les fichiers distribués, read 2026-10-05
No audit report sits in the repository either, and this absence was searched for rather than assumed: the full tree, recursive, came back with zero paths containing the word "audit" apart from a dependency-scanning configuration file that is not a security review.
Source near/nearcore — arborescence complète, read 2026-10-05
Is anyone still working on it?
Yes: 296 commits from 24 distinct Git identities in the 90 days we measured, the most recent landing the same day we checked. The project also ships numbered releases rather than leaving its history to commits alone, 100 of them at the time we counted.
That count comes from walking the commit list for the window itself, not from a dashboard summary, and it sits in the same range as the busiest repository on this site, which counted its activity the same way: by paging through the actual objects GitHub returns, because a ceiling number ending in a round hundred is a sign of pagination, not of truth.
Twenty-four Git identities are not twenty-four people either way: one contributor can sign commits under several names, and the number says how many signatures showed up, not how many engineers.
Source GitHub, read 2026-10-05
What shows it is running
- The chain responds — block height 218678650, 100 active validators
- The repository is actively worked on — 296 commits in the 90 days to 2026-10-05, 24 distinct Git identities
Source rpc.mainnet.near.org · GitHub, read 2026-10-05
- The core repository is made public
- Maximum inflation cut from 5% to 2.5%
- Start of the 90-day window we measured
- This page, checked and published
Where does it actually run?
On its own chain, which we checked responds: a public node answered with the current block height and a hundred active validators in this epoch. Unlike a network of nodes that reports outside data onto other chains, NEAR settles its own transactions end to end.
That distinction matters because several projects on this site run entirely on top of someone else's ledger. A network of oracle nodes exists to carry data onto other chains and settles nothing of its own. NEAR is the chain being settled on, not a service riding across one.
Source rpc.mainnet.near.org, read 2026-10-05
NEARNetwork verified by reading its current block height on a public node.
What the token actually does
NEAR pays for every transaction and is staked by validators to help order blocks. There is no supply cap: well over a billion tokens already circulate, and a protocol change in October 2025 halved the maximum rate at which new ones can still appear.
1,308,160,559 NEAR circulate out of 1,308,160,513issued, a gap narrow enough that most of what exists is already in someone's hands rather than waiting in a reserve. No company or foundation issues it the way the oldest chain on this site has no issuer at all; here a named foundation exists, and its own site names two people as founders, which is already more than several other projects we have checked are willing to say.
The gas fee itself splits in two, read directly in the running code rather than guessed from a blog post: 70% of every fee is destroyed outright, and the remaining 30% goes to the account that executed the call.
Source near/nearcore — core/parameters/res/runtime_configs/parameters.yaml, read 2026-10-05
A separate change, dated and versioned rather than announced, cut the ceiling on new issuance from 5% to 2.5% a year, effective with version 2.9.0 on 2025-10-21.
Source near/nearcore — CHANGELOG.md, read 2026-10-05
- In circulation1,308,160,559
- What the token is forDirectly verified
- How many existThird-party supported
- How new ones appearDirectly verified
- Who can change the rulesNot verified
The KLEAR3 rating
3.8out of 5 in the KLEAR3 rating
Each axis gets 0 to 5 from the facts in the register, against thresholds written in advance. The overall score is their weighted average, and it only appears once at least 5 of the 6 axes rest on a measured fact.
- Audits ×1.3
- Open code ×1.2
- Business model ×1.1
- Token alignment ×1
- Governance ×1
- Maturity ×0.7
- Open code5
public repository · MIT OR Apache-2.0 licence · 100 published releases · 296 commits in 90 days
- Audits3
code reviewed in the open: 296 commits in 90 days, 8 years in production
What holds it back no report in our registry · no report opened: we do not know what they found
- Business model5
business model published by the protocol · paid by whoever sends the transaction · collected by the account that executed the call · seventy percent of the gas spent is destroyed on every transaction; the remaining thirty percent goes to the account that executed the call
- Governance4
decided by a public proposal process (NEP), with no vote · 51 improvement proposals published · process open for 7 years
- Token alignment2
supply measured, with no cap: nothing bounds issuance · circulating share known · emission schedule published · terminal inflation 2.5% a year
What holds it back fate of the initial allocations not established
- Maturity4
8 years of public code · 1 network verified by us
We rate the protocol: how it is built, how much of it can be checked, how its token lines up. Never an investment. A high score is not advice to buy, and a low one is not advice to sell.
What we could not establish
Two things, and both matter to a reader deciding what to trust. No audit report of the protocol itself exists anywhere we looked, searched rather than assumed, and no on-chain vote weighted by the token decides protocol changes: the public proposal process runs independently of who holds what.
That second point is easy to misread as a flaw. A public proposal process with 51 documents published and 307 contributions merged since 2019-06-05 is a real, open mechanism. It is simply not the same thing as a token vote, and this page does not blur the two.
Source near/nearcore — arborescence complète · near/NEPs, read 2026-10-05
Frequently asked questions
Is NEAR open source if GitHub says GPL-3.0?
Its own code is not GPL-licensed at all, and GitHub's badge is simply wrong here. The project's build configuration states MIT OR Apache-2.0, and every licence file physically present in the repository, dozens of them, matches that statement rather than the badge.
Walking the full repository tree settles it: of roughly 3000 paths, not one contains GPL text, while LICENSE-MIT and LICENSE-APACHE files sit inside well over 50 separate crates. The files disagree with the badge, not with each other.
GitHub's license detector reads a repository and guesses from what it finds; it is not a legal reading of the code, and it can be wrong. Here it was, confidently and in the direction most likely to scare off a reader. A directory that prints the badge without opening a single file is repeating that same mistake, at the cost of a project that actually ships the most permissive pairing on this entire list.
Does NEAR have an audit?
Not of the protocol itself, as far as a full search of its own repository shows. The tree was read recursively, close to 3000 paths, and none of them is an audit report. The only match for the word is a dependency-scanning configuration file, not a security review.
That is an absence we searched for rather than assumed, which matters: a project never looked at carries a different risk than one reviewed and cleared, and this page does not conflate the two, even though both can sound just as reassuring in a pitch deck.
Nothing here says an audit does not exist in a place we did not check, and nothing here says the code is unsafe. It says what a full, recursive read of the public repository actually returned on the day we read it.
Can NEAR holders vote on protocol changes?
Not by holding the token. Changes go through a public proposal process on GitHub, open since 2019, with 51 proposals published and 307 contributions merged into it. None of that depends on how many tokens a contributor holds, which is a different model from a token-weighted vote.
The useful comparison is not whether this is better or worse than a vote, it is what each model actually protects against. A public process with a long paper trail is hard to game, even without a ballot attached to it, and that is worth saying plainly rather than implying.
What it does not establish is who ultimately merges a contested proposal, and this page has not opened that question. 51 documents and a long-running repository describe a process that exists and runs; they do not describe who holds the final word.
Related: our rollup explainer covers the scaling approach NEAR’s own sharding competes against rather than uses, and our Arweave profile shows a chain built around a single guarantee, permanent storage, instead of general-purpose contracts.
Ask an assistant to summarise it and check what it claims.