bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 3쪽 중 3쪽
epresented in the
same financial state while retaining distinct origins and verification models.
8 Artificial intelligence and agentic finance
8.1 A programmable financial environment for agents
Injective’s architecture is naturally suited to autonomous software because financial actions are
exposed through stable module interfaces, market state is public, and settlement is deterministic.
An agent can discover available functions, query balances and markets, construct transactions,
and request signatures against the same infrastructure used by human-facing applications, without
receiving any special authority from consensus.
Current developer tooling demonstrates this model by exposing balances, markets, orders, trans-
fers, and other workflows through command-line interfaces, Model Context Protocol (MCP) servers,
and reusable agent skills. These tools remain clients of the protocol rather than components of con-
sensus: transaction authority stays in the signing layer, allowing an agent to research and prepare
broadly while holding only the execution authority assigned to it.16,17
Within this broader toolchain, MCP provides a standardized interface through which compatible
agents can discover and invoke Injective capabilities. The main Injective MCP server exposes chain
queries and transaction workflows, including market data, orders, transfers, bridging, and EVM
transaction submission. A separate documentation server provides semantic search and retrieval
across Injective’s developer materials. Together, they give agents both an execution surface and a
current source of protocol context without defining the agent’s reasoning process or expanding its
authority.16,17
The resulting architecture separates reasoning, authorization, and execution. A reasoning layer
selects an action, a policy and signing layer determines whether that action may be authorized, and
the chain applies deterministic transaction validity, market checks, and settlement. Autonomous
systems can therefore operate against a deterministic financial core without placing probabilistic
reasoning inside consensus.
18 INJECTIVE WHITEPAPER OCTOBER 2026
8.2 Machine payments through x402
x402 embeds payment into an HTTP request cycle using the 402 Payment Required status. A
client that requests a protected resource receives payment requirements, signs an authorization, and
the signed authorization is relayed to a facilitator, which sends the onchain transaction and waits for
settlement before the client retries the request with proof of payment. Injective’s implementation
uses USDC on the EVM path.18
This protocol turns a web endpoint into a metered service without a conventional account or
subscription. It is especially useful for machine-to-machine access to data, models, or compute, and
it gives agents a payment path that can be composed directly with onchain USDC settlement.
8.3 Bounded authority and deterministic automation
Agent authority can be bounded through narrow keys or smart accounts, asset and amount
limits, market allowlists, price-impact limits, simulation, approvals, rate limits, and execution logs.
Because these controls are enforced outside the reasoning model, they can remain deterministic even
when the agent’s planning process is probabilistic.
Scheduled contracts offer a complementary model by executing deterministic logic at defined
block stages without relying on a language model. Agents are well suited to interpretation and
planning, whereas chain-scheduled contracts are better suited to invariant execution; combining
them can keep discretionary reasoning outside the consensus boundary while preserving determin-
istic rules for authorization and settlement.
9 The INJ token economy
9.1 Network functions
INJ is the native asset used for proof-of-stake security, governance, and network fees. Validators
bond INJ directly or receive delegated stake, and the resulting voting power determines participation
in consensus; governance uses staked participation to decide proposals and parameter changes, while
transaction fees compensate the network for computation and state use.
The token also connects network security and financial activity to supply through two sepa-
rate mechanisms: dynamic protocol issuance and recurring burns. The separate paperINJ: A
Programmable Token Economy for Deflationary Accelerationprovides the dedicated historical and
economic treatment, while this section focuses on the mechanisms directly relevant to the protocol
architecture.19
19 INJECTIVE WHITEPAPER OCTOBER 2026
9.2 Moving Change Rate Mechanism
The mint module adjusts the annualized supply rate block by block according to the difference
between the actual bonded proportion and a target. LetSRbe the current supply rate,SRC
the supply-rate-change parameter,Bthe configured blocks per year,BP actual the actual bonded
percentage,BP goal the target, andSRlower andSR upper the approved bounds. The Moving Change
Rate Mechanism is:19
SRnew =
⎧
⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎩
min
(︃
SR+ (BPgoal −BP actual)SRCB , SRupper
)︃
BPactual < BPgoal,
SR BP actual =BP goal,
max
(︃
SR− (BPactual −BP goal)SRCB , SRlower
)︃
BPactual > BPgoal.
(8)
Equation 8. Moving Change Rate Mechanism
When bonded participation falls below the target, the controller raises the supply rate toward
the upper bound; when participation rises above the target, it lowers the rate toward the lower
bound. At the target, the rate remains unchanged, making the mechanism a feedback controller for
staking participation rather than a fixed emission schedule.
Governance proposal 617, passed in January 2026, set the annual issuance bounds to 2.2% and
4.4%, retained a 60% goal-bonded ratio and 50% rate-change parameter, and set blocks per year
to 63,072,000. These were the active parameters documented for this October 2026 edition of the
Injective Whitepaper.20
9.3 Community BuyBack
The Community BuyBack is a recurring onchain process in which participants commit INJ for a
pro rata share of contributed ecosystem revenue, with the committed INJ then burned. The program
replaced the earlier winner-take-all burn auction in October 2025, broadening participation while
preserving permanent token removal.
Under the exchange module’s fee design, 40% of positive trading fees is paid to the designated fee
recipient, whiletheremaining60%accruestotheprotocolrevenuepoolthatsuppliestheCommunity
BuyBack. Other applications and participants can also contribute revenue. This makes the burn
mechanism horizontally scalable: as the application ecosystem expands and generates more revenue,
the potential buyback pool can grow without requiring base-layer transaction fees or congestion to
rise.
Dynamic issuance and Community BuyBack burns remain separate flows: the first adapts stak-
ing incentives to bonded participation, while the second converts contributed ecosystem value into
token removal.2,19
20 INJECTIVE WHITEPAPER OCTOBER 2026
10 Governance
Injective governance coordinates software upgrades, module parameters, market settings, com-
munity spending, and other state changes authorized by the protocol. Staked INJ supplies voting
power, and proposals pass or fail under configured quorum, threshold, and veto conditions.
Thisprocesshascarriedthenetworkfromitsearlyexchange-focusedarchitectureintoaMultiVM
financial platform. Consensus configuration, EVM activation, module upgrades, oracle providers,
scheduled-contract registrations, and token-economic parameters can all be coordinated through the
same onchain mechanism.
Protocol governance and asset administration nevertheless remain distinct. Chain governance
controls shared protocol state, while a TokenFactory administrator controls its namespace, a
permissioned-asset issuer controls retained roles, Circle controls native USDC issuance and CCTP
attestations, and each application controls its own contracts and interface. The separation allows
the base protocol to remain common even when product and asset policies are independently
configured.
11 Security
11.1 Network security
Network security rests on proof-of-stake consensus. Validators commit blocks through the
greater-than-two-thirds voting-power threshold formalized in Equation 2, while validators and del-
egators are economically aligned through staking rewards and slashing. Deterministic finality gives
each committed state transition one authoritative result.2,4
11.2 Financial security
Security extends into the financial layer through several coordinated mechanisms. Frequent
Batch Auctions reduce the continuous time-priority race at matching. Oracle medians and cumu-
lative price state support deterministic marks and time-weighted pricing. Initial and maintenance
margin constrain leverage, reduce-only liquidations close impaired positions, and available insurance
capital may absorb eligible residual shortfalls.
The security model is layered because consensus finalizes state but does not originate every
input. Oracle modules govern how external prices enter protocol state; exchange modules apply
market and risk parameters; IBC, Peggy, and CCTP validate external events under different veri-
fication models; governance and asset administrators act within their respective scopes; and agents
operate through independent signing and authorization controls. These components converge on
one finalized ledger, but they do not collapse into one trust model. Finality makes an accepted state
21 INJECTIVE WHITEPAPER OCTOBER 2026
transition authoritative, while the validity of its inputs and authorizations remains governed by the
subsystem that produced them.
11.3 Software assurance
Because native modules serve many applications, protocol correctness has system-wide value.
Deterministic tests, simulation, property and invariant testing, audits, controlled upgrades, bug
bounties, and monitoring form the assurance process around modules, VM integrations, contracts,
and MultiVM mappings. Role separation, namespace sealing, and transparent governance add
complementary controls for administrative state.
Table 3: Layered security mechanisms
Layer Core mechanism Protected transition Complementary controls
Consensus Greater-than-two-thirds
precommit threshold
Canonical block
commitment
Staking, slashing, validator
operations, monitoring
Oracle Signed reports, medians,
cumulative price state
External data to
protocol price
Provider diversity, cadence,
bounds, TWAP
Market risk Margin checks and
reduce-only liquidation
Position state to
orderly close
Parameters, liquidity,
available insurance capital
Software Deterministic module and
contract execution
Valid transaction to
state update
Tests, audits, invariants,
staged upgrades, bounties
Settlement Light clients, attestations,
and bridge verification
External event to
Injective asset state
Asset provenance and
path-specific limits
Administration Governance and
namespaced role policy
Authorized parameter
or asset action
Role separation, sealing,
multisignature controls
Agents Policy-controlled signing
authority
Machine intent to
signed transaction
Simulation, approvals, limits,
and execution logs
22 INJECTIVE WHITEPAPER OCTOBER 2026
12 Conclusion
Injective is a Layer-1 blockchain whose protocol implements the common machinery of a financial
system. Consensus establishes deterministic finality; the exchange module converts orders into a
shared market transition; oracle and insurance modules support pricing and loss absorption; and
TokenFactory and Permissions provide native issuance and asset policy. MTS gives EVM and
WASM one canonical asset state, while IBC, CCTP, and bridge infrastructure connect external
assets under distinct verification models.
Its defining choice is to provide this common machinery once, at the protocol layer. Applications
built for EVM, WASM, or native module interfaces can operate using the same unified assets,
orderbooks, risk state, and settlement, even when they present very different products to users.
The onchain orderbooks remain protocol infrastructure, while liquidity comes from participants
that quote into them.
The 2018 whitepaper set out to build an exchange protocol that was trustless, publicly verifiable,
liquidity neutral, and resistant to front-running. The live network now realizes those objectives
at the base layer and extends them through a broader system of native markets, programmable
execution, asset issuance, interoperability, and automated finance. What began as an exchange
settlement protocol has become a sovereign state machine for programmable markets, open to
developers, institutions, and autonomous agents alike.
23 INJECTIVE WHITEPAPER OCTOBER 2026
References
[1] Eric Chen and Albert Chon, “Injective Protocol: A Collision Resistant Decentralized Exchange
Protocol,” December 2018.
[2] Injective Foundation, “Injective Core,” public source baseline v1.20.3-safeharbor.2, GitHub
commit de8a811.
[3] Remco Bloemen, Leonid Logvinov, and Jacob Evans, “EIP-712: Typed Structured Data Hash-
ing and Signing,” Ethereum Improvement Proposals, https://eips.ethereum.org/EIPS/eip-712.
[4] Ethan Buchman, Jae Kwon, and Zarko Milosevic, “The Latest Gossip on BFT Consensus,”
2018, https://arxiv.org/abs/1807.04938.
[5] Petar Zhivkov and Eric Chen, “Dynamic Fee Markets at Sub-Second Timescales: Adapting
EIP-1559 for High-Throughput Blockchains,” Mathematics 14, no. 7 (2026): 1232,
https://doi.org/10.3390/math14071232.
[6] Tim Roughgarden, “Transaction Fee Mechanism Design,” Journal of the ACM 71, no. 4 (2024):
Article 30, https://doi.org/10.1145/3674143.
[7] Eric Budish, Peter Cramton, and John Shim, “The High-Frequency Trading Arms Race: Fre-
quent Batch Auctions as a Market Design Response,” Quarterly Journal of Economics 130, no.
4 (2015): 1547-1621, https://doi.org/10.1093/qje/qjv027.
[8] Eric M. Aldrich and Kristian Lopez Vargas, “Experiments in High-Frequency Trading: Com-
paring Two Market Institutions,” Experimental Economics 23, no. 2 (2020): 322-352,
https://doi.org/10.1007/s10683-019-09605-2.
[9] Hamda Al-Breiki, Muhammad Habib ur Rehman, Khaled Salah, and Davor Svetinovic, “Trust-
worthy Blockchain Oracles: Review, Comparison, and Open Research Challenges,” IEEE Ac-
cess 8 (2020): 85675-85685, https://doi.org/10.1109/ACCESS.2020.2992698.
[10] LorenzBreidenbach, ChristianCachin, AlexCoventry, AriJuels, andAndrewMiller, “Chainlink
Off-chain Reporting Protocol,” version 1.2, February 24, 2021,
https://research.chain.link/ocr.pdf.
[11] Damien Ackerer, Julien Hugonnier, and Urban Jermann, “Perpetual Futures Pricing,” Mathe-
matical Finance 36, no. 3 (2026): 481-499, https://doi.org/10.1111/mafi.70018.
[12] Ravi S. Sandhu, Edward J. Coyne, Hal L. Feinstein, and Charles E. Youman, “Role-Based
Access Control Models,” Computer 29, no. 2 (1996): 38-47, https://doi.org/10.1109/2.485845.
[13] Siddhant Shah and Eugene Pinsky, “Perpetual Futures in Decentralised Finance: Mechanics,
Economic Claims, and the Drivers of Trading Volume,” International Journal of Financial
Studies 14, no. 7 (2026): 178, https://doi.org/10.3390/ijfs14070178.
24 INJECTIVE WHITEPAPER OCTOBER 2026
[14] Christopher Goes, “The Interblockchain Communication Protocol: An Overview,” 2020; Cos-
mos, “Interchain Standards.”
[15] Circle Internet Financial, “Cross-Chain Transfer Protocol V2.”
[16] Model Context Protocol, “Model Context Protocol Specification,” 2026-07-28.
[17] Injective Labs, “Injective MCP Server”; “Injective Agent SDK.”
[18] x402 Foundation, “x402 Specification”; “Exact Scheme on EVM.”
[19] Injective Research, “INJ: A Programmable Token Economy for Deflationary Acceleration,” May
2024, https://injective.com/INJ_Tokenomics_Paper.pdf.
[20] Injective governance proposal 617, “The INJ Supply Squeeze,” onchain proposal, MsgUp-
dateParams, and final tally, January 2026, https://injhub.com/proposal/617/.프로젝트로 돌아가기