PaperlayerView exact revision on arXiv

Source-linked AI summary

TessIndex: Capability Verified Identity System for the Agent Economy

Mehul Goenka, Tejas Pathak, Siddharth Asthana

arXiv:2608.21942v1cs.AIcs.CRcs.MAcs.NI

TL;DR

The agent economy lacks persistent identity, verifiable capability evidence, and links among creators, performance, and project value. TessIndex addresses these gaps through a dual-plane identity system that combines blockchain commitments with centralized metadata, yielding verified primitives connected to discovery, commerce, reputation, and settlement.

  • Problem

    Agentic workflows lack identity infrastructure for accountability, capability claims lack execution evidence, and creator, performance, and project value remain disconnected.

  • Method

    TessIndex combines blockchain-anchored commitments for identity, ownership, and verification with centralized metadata for discovery, commerce, and reputation.

  • Results

    TessIndex establishes verifiable links among operational metadata, payment identity, and immutable on-chain identity, with settlement conditioned on predicate verification.

  • Takeaways & Limitations

    TessIndex integrates persistent identity, verified capabilities, execution, reputation, and economic workflows into a trust layer for agentic primitives.

Abstract

from arXiv · show

Software systems have traditionally been organized around applications where human users act as principal decision-makers. Recent developments in agentic capabilities alter this paradigm: software agents now autonomously translate high-level goals into structured tasks, orchestrating tools, services and sub-agents to execute complex workflows. This evolution gives rise to an agent economy where these autonomous agents capture real economic value. However, the infrastructure required to support the agent economy fails across three critical dimensions: the absence of persistent identity infrastructure prevents systemic accountability in agentic workflows; capability claims remain self-declared not backed by verifiable execution evidence; and the disconnect between creator identities, agent performance, and project value hinders the economic valuation of agents as assets. While existing registries provide naming and discovery, unifying these features around a persistent identity anchor remains largely unaddressed. TessIndex is a capability-verified identity system for agent primitives that utilizes a dual-plane architecture: the blockchain records compact commitments for identity, ownership, and verification, while centralized servers maintain dynamic metadata for discovery, commerce, and reputation. It establishes: persistent identities across agent primitives to enforce systemic accountability in autonomous workflows; a predicate-based verification process replacing self-declared claims with cryptographic capability proof; an identity infrastructure that links agent performance to both project and creator identities while capturing value through tokenization. Ultimately, TessIndex serves as an integrated infrastructure that binds an agent's existence across capabilities, execution, and reputation into a single persistent identity.

I. INTRODUCTION

The agent economy requires shared infrastructure that connects autonomous actions to authority, verified capabilities, and economic value. TessIndex addresses these gaps with a unified identity, verification, registry, and interoperability framework for agent economy primitives.

  • Autonomous agents coordinate tools, services, skills, and interfaces to execute workflows that produce economic value.
  • Current infrastructure lacks accountability linking human intent and authority to actions performed by agents and invoked primitives.
  • Capability claims remain self-asserted without systematic verification or concrete execution evidence under declared operating conditions.
  • Existing identity and registry efforts emphasize naming, discovery, and secure resolution but generally do not cover tools, services, commerce, and value.
  • Contributions: TessIndex contributes persistent identities, cryptographic capability proofs, rich commercial and performance metadata, and interoperability across registries and protocols.

A. Unified Identity System for Primitives in Agent Economy

TessIndex defines a unified identity infrastructure for agents, tools, services, and other primitives, combining persistent identity with verification, discovery, interoperability, and lifecycle governance. Its dual-plane architecture anchors security-critical commitments on blockchain while centralized services manage richer dynamic metadata.

  • Unified Identity System for Primitives in Agent Economy: Agent economy primitives are distributed across disconnected identity systems, motivating a common representation for consistent ecosystem participation.
  • Persistent identity lets users verify a primitive and its creator while supporting accumulated reputation, transaction history, and commercial credibility.
  • Capability verification anchors tamper-resistant verification receipts so claimed functions and security properties have verifiable proof.
  • Intent-based discovery uses rich metadata to select agents, tools, or services by context, capability, interface, and verification status.
  • TessIndex is designed to interoperate with existing registries, protocols, and discovery systems rather than operating in isolation.
  • Dual-Plane Architecture: The blockchain stores compact commitments and immutable records, while centralized servers retain flexible, scalable, and expressive metadata.

V. KEY DOMAINS

TessIndex’s key domains define what is indexed, how identities are classified and controlled, and how records connect across ecosystems. The identity structure links primitives to owners, projects, contracts, and tokens for lifecycle management and assetization.

  • Identity: TessIndex defines an identity domain covering indexed entities, stable identifiers, ownership authority, naming, and ecosystem interoperability.Its identity model anchors records through canonical identifiers and specifies who may authorize updates.
  • Objects: The registry supports agents, tools, services, skills, channels, projects, and agentic applications as distinct object types.These components participate in discovery, invocation, communication, or collective project organization.
  • Lifecycle: TessIndex preserves lifecycle semantics through states including unverified, inactive, active, verified, and revoked.These states maintain discovery meaning as records change over time.
  • Identifiers: TessIndex assigns each agent a unique 96 bit cryptographically secure random agent_id encoded in Crockford Base32.The agent_id functions as the central identity anchor across the ecosystem and identifier surfaces.
  • Identity Structure: Agent identities may stand alone or belong to projects, with Agent Identity Tokens, Agent Contracts, and Agent Project Tokens linking ownership, verification, and assetization.Projects reference creator identities on chain, while tokens may support trading against stablecoin cryptocurrencies based on agent performance and utility metrics.

3) Schema

TessIndex combines a strict core schema with optional extensions, rich capability descriptions, commerce fields, and assetization metadata. This schema supports interoperable discovery, deterministic verification, paid execution, and performance-based analysis.

  • Schema: TessIndex defines a structural contract with a small required baseline and an extensible layer for richer discovery and operations.Optional fields can be ignored by clients without breaking the stable core schema.
  • Core Fields: The core schema spans Identity, Control, Interoperability, and Commerce domains, with fields for identity, authorization, ecosystem integration, and commercial primitives.These fields allow resolvers, orchestrators, and external integrations to interpret records consistently.
  • Schema Richness: Rich schema fields such as descriptions, skills, capabilities, use cases, and status improve semantic retrieval, ranking, compatibility checks, and cross-registry normalization.The schema harmonizes fields commonly used across indexing services while retaining extensibility.
  • Capability Definition: Capabilities require explicit interfaces, constraints, and representative subtasks so verification and orchestrator-driven execution can be performed deterministically.Supported input and output formats include JSON, text, and image representations.
  • Commerce and Assetization: Commerce metadata includes pricing, payment rails, settlement modes, wallet addresses, service details, and assetization references linked to performance telemetry.Tracked performance data includes latency, throughput, tokens used, workflow counts, average success rate, and user ratings.

B. Runtime

The Runtime domain operates TessIndex as a control layer for naming, verification, registration, protocol adaptation, and lifecycle management. Its predicate-based verification combines execution proofs with trusted runtime attestations before granting Verified status.

  • Runtime: Runtime components assign identity, verify capabilities, adapt protocols, provision wallets, and register primitives across blockchain and centralized servers.Composable microservices allow these functions to evolve independently across heterogeneous entities and execution environments.
  • Naming: The Primitive Naming Service issues agent_id and root_domain values while maintaining their unique mapping to registry records.Domain resolution uses root and authoritative naming layers to map domains to canonical records and endpoints.
  • Verification: Capability verification assigns a representative task, collects execution proof objects, and validates each object against its referenced predicate_id.Agent Contracts specify predicate identifiers, while the on-chain predicate library stores proof schemas and verification mechanisms.
  • Verification: Trusted Execution Environment attestations and proof receipts are verified on chain by BFT consensus as part of the onboarding decision.Successful verification of both capability proofs and runtime integrity attestations grants the agent Verified status.
  • Protocol Adaptation: The wrapping service maps native endpoints to A2A, x402, and AP2 interaction patterns without altering the agent’s internal logic.A sidecar enables standardized invocation, pricing discovery, and payment-compatible execution.
  • Registration: Registration persists the manifest, creates a wallet, and mints an Agent Identity Token that stores or references the manifest hash.The resulting linkage connects operational metadata, payment identity, and immutable on-chain identity.

2) Data Model

TessIndex separates immutable identity commitments from mutable operational data while binding manifests, ownership, metadata, and verification records cryptographically. Its orchestration layer routes verified execution under policy and context constraints.

  • Storage Split: TessIndex stores compact identity records on blockchain and richer manifests in centralized servers according to differing trust, cost, update, and operational requirements.The blockchain serves as the authoritative trust anchor for descriptive and operational data held centrally.
  • Where Identity Lives: Primitive identifiers and ownership bindings remain persistent on chain, while metadata, endpoints, policies, and operational configuration can evolve off chain.Cryptographic commitments preserve correspondence between changing details and the registered identity.
  • Data Integrity: Manifest hashes, creator signatures, ownership checks, version commitments, and linked verification records protect data integrity and traceability.Consumers can independently check whether registry data is current, authentic, and consistent with the identity record.
  • Kernel Architecture: The orchestration kernel decomposes user intent into DAG-based plans, selects primitives using identity and capability metadata, and enforces policy and verification during execution.It coordinates intent interpretation, routing, context storage, state transitions, and execution checks.
  • Memory Architecture: TessIndex separates personalized knowledge, durable cross-session context, and temporary interaction state across three memory structures.These are the Personalized Knowledge Graph, Long Term Memory, and Short Term Memory.
  • Execution and Verification: Each system call enters a scheduler queue, reaches an executor, produces telemetry, and passes through predicate verification before being treated as valid.The execution surface includes endpoints, schemas, executor bindings, verification hooks, authentication, authorization, and security constraints.

4) Lifecycle

TessIndex uses lifecycle controls to preserve primitive identity continuity while supporting registration, renewal, status changes, versioning, and revocation. Authenticated record authority ties updates to explicit controllers and maintains traceability over time.

  • Lifecycle: Lifecycle management preserves identity continuity while allowing primitives to evolve, expire, renew, suspend, or exit active registry use.The lifecycle covers registration, renewal, status, versioning, and revocation.
  • Registration: Registration requires signed metadata, capability, endpoint, ownership, and verification information before blockchain and centralized records are created.The registry verifies the signature, assigns an identifier, records an identity commitment on-chain, and stores the richer manifest centrally.
  • Renewal: Renewal refreshes ownership, metadata, endpoints, attestations, policies, and verification records, with missed deadlines potentially marking a primitive expired or inactive.Historical identity records remain preserved when renewal does not occur.
  • Versioning: Versioned commitments track changes to manifests, capabilities, endpoints, policies, predicates, execution surfaces, and verification records without changing the stable primitive identifier.Historical versions support auditability while the current version remains available for discovery and orchestration.
  • Revocation: Revocation removes or deactivates primitives or claims from active discovery and orchestration while retaining historical records for auditability, disputes, and provenance.Owner, registry, or verifier actions can trigger revocation, including validated reports and security failures.
  • Record Authority: Record mutations are restricted to the creator_id bound to the Agent Identity Token and authenticated through creator signatures and wallet authorization.The update path uses Ed25519 signing, public-key verification, and EIP 712 wallet authorization.

3) Endpoint Resolution

TessIndex resolves callable endpoints by combining execution constraints with adaptive routing policies and live or historical telemetry. It retries failures, switches endpoints, and updates health signals to preserve execution continuity.

  • Endpoint Records: Endpoint records can include primary and secondary URLs, success rates, p95 latency, timestamps, and bearer authentication references.The supplied example shows a primary endpoint with 0.995 success rate and 420 ms p95 latency, alongside a secondary endpoint with 0.990 and 510 ms.
  • Resolution Logic: The adaptive endpoint resolver validates authentication and selects endpoints using context-dependent policies over uptime, success rate, latency, and cost.Creators choose the routing policy at deployment time.
  • Routing Policies: Supported policies include round robin, least connections, latency-based selection, and weighted routing.Weighted routing uses creator-assigned weights, while latency-based selection favors the endpoint with the maximum resulting score when latency dominates.
  • Metrics: Endpoint ranking combines normalized uptime, success rate, latency, and cost into a context-weighted score.Higher normalized latency and cost represent worse performance or expense, while the weights are context dependent.
  • Failure Handling: The resolver retries failed requests, switches to alternate endpoints, and updates health and ranking signals to avoid repeatedly selecting degraded endpoints.This behavior is intended to support graceful degradation and rapid recovery when endpoints become unavailable.

4) DevEx and Operations

TessIndex combines creator-facing publication workflows, automated updates, read-oriented performance design, and interoperability across communication and commerce protocols. Its operational model supports publication, verification, registration, tokenization, and cross-ecosystem integration.

  • Registration UX: Creators can publish through the Registration Wizard UI or Tesseris CLI by submitting a repository link, environment variables, and agent metadata.The backend handles deployment, capability verification, registration, and tokenization.
  • Automation: Repository webhooks and sequential pipelines automate deployment, Agent Contract creation, capability verification, registration, and on-chain and off-chain updates.Creator wallet signature verification triggers the update flow through the Registration Service.
  • Performance Characteristics: Tiered caches and a lean index layer target predictable performance for read-heavy discovery and filtered retrieval.Query-result and endpoint-record caches use explicit freshness windows.
  • Cost Model: The cost model uses an ongoing creator subscription fee and a specialized creator launch fee for agent token launch flows.The launch fee supports complex actions such as bonding curve operations.
  • Interoperability: The interoperability layer normalizes external ecosystems through adapters and source-linked metadata while preserving provenance and trust context.It supports unified discovery and invocation across heterogeneous protocols and registries.
  • Communication Protocols: TessIndex supports A2A for agents and MCP for tools as standard communication protocols.These protocols define the supported communication surface for agent invocation and tool interaction.
  • Commerce Protocols: For agentic commerce, TessIndex supports x402 and AP2 with explicit user authorization and auditable payment rails.The system describes an end-to-end transaction cycle across these commerce protocols.

E. Economy

TessIndex provides a unified economy-facing discovery surface for agents, services, and endpoints. It narrows candidates with structured filters, ranks them by intent relevance, and separates fast retrieval from adaptive endpoint selection.

  • Ecosystems for Discovery: TessIndex supports discovery across federated ecosystems by normalizing heterogeneous schemas and returning agents and services in a consistent form.Supported surfaces include Coinbase x402 Bazaar, Ethereum ERC 8004, Virtuals, Fetch.ai, HOL, NANDA Index, and Google discovery surfaces.
  • Discovery Mode: The system offers name-based discovery and orchestrator-driven discovery from natural-language user intent.Orchestrators can retrieve, rank, and route to suitable agents automatically.
  • Query Interface: A semantic registry API lets orchestrators express intent, capability, trust, interface, and policy constraints instead of requiring a known agent identifier.The API returns ranked or filtered candidate sets for downstream selection.
  • Filters: Structured filters narrow candidates by services, capabilities, endpoint types, protocols, lifecycle and verification status, payment rails, and price range before ranking.Deterministic filtering reduces the candidate space prior to ordering.
  • Ranking Mechanism: Candidate ranking uses a two-stage pipeline in which LambdaMART creates an initial order and a cross encoder performs deeper semantic reranking.The final list reflects similarity between user intent and candidate metadata.
  • Retrieval Mechanism and Speed: A lean registry retrieves semantically similar candidates quickly while the Adaptive Endpoint Resolver handles reliable endpoint choice separately.This separation supports low-latency discovery without removing adaptive execution logic.

2) Agent Wallet Integration

TessIndex links agent wallets to persistent identities so ownership, settlement, and payments remain attributable and verifiable. Its payment design supports interoperable, auditable, conditional settlement across rails and chains.

  • Wallet Identity: Wallet binding cryptographically ties payment-relevant wallets to agent identity anchors, supporting verifiable payments, ownership, assetization, tokenization, and revenue flows.The system uses creator-controlled derived agentic wallets and on-chain Agent Identity Tokens to link creator wallets to agent identities.
  • Payment Protocols: TessIndex supports x402 and AP2 through user-agent and facilitator-agent architectures that capture immutable mandates for delegated authority and dispute resolution.The facilitator coordinates execution while the user agent signs transactions with explicit user authorization.
  • Settlement: Conditional settlement holds user funds in escrow and releases them only after a quorum-signed Proof of Task Execution reaches the escrow.This implements a verify-then-pay model for agentic payments.
  • Auditability: TessIndex maintains off-chain payment audit trails with on-chain anchors, retaining mandates, transaction hashes, proof artifacts, PoTE receipts, and validator metadata.These records support reconstructible payment paths, auditability, dispute resolution, and payment traceability.
  • Interoperability: TessIndex is chain agnostic at discovery and settlement layers, separating control from independent per-chain payment rails, escrow contracts, and wallet accounts.Price discovery can be user-led or orchestrator-driven through the agent’s /price endpoint, with negotiation supported for variable pricing.

F. Trust and Security Layer

TessIndex’s Trust and Security Layer combines identity binding, execution evidence, consensus, and settlement controls to make agentic interactions verifiable and auditable. Trust operates across identity, execution, and settlement levels.

  • Layer Overview: The Trust and Security Layer establishes cryptographic, operational, and settlement guarantees while preserving evidence for auditability and dispute resolution.It verifies identity, validates execution, controls settlement, and captures trust signals across reliable agentic interactions.
  • Identity: Creators sign primitive manifests with delegated Ed25519 keys, whose public keys bind to creator wallets through on-chain delegation records and minted Agent Identity Tokens.The resulting AIT associates the agent identity with an immutable on-chain record tied to agent_id.
  • Execution: Every primitive action produces a cryptographically verifiable execution receipt that third-party verifiers validate against the corresponding predicate.Verifier consensus anchors final receipts on chain, while workflow receipts form a graph whose Merkle root represents workflow-level PoTE.
  • Settlement: Escrow progresses through defined states and releases funds only after Byzantine fault tolerant consensus confirms predicate-verified task execution.No settlement occurs without successful predicate verification.
  • Trust Signals: Mandates, verifier signatures, request signatures, TLS receipts, deterministic hashes, staking, and randomized verifier selection provide layered trust signals and accountability.These mechanisms cover authorization, external interactions, consensus, communication, tamper resistance, and collusion risk.

2) Reputation

TessIndex models reputation as a layered, continuously updated trust signal for verifiers and agents. It combines execution telemetry, verification records, user feedback, and access-control mechanisms to support ranking and routing.

  • Reputation Model: TessIndex separates verifier reputation from agent reputation, using historical reliability for verifier eligibility and completed-task performance for agent assessment.Agent reputation can include success rate, failure rate, latency, throughput, user feedback, and other execution indicators.
  • Trust and Integrity: These metrics distinguish agents that merely claim capabilities from agents that demonstrate reliable performance through repeated execution.The system also uses predicate enforcement, layered access control, and abuse-resistance mechanisms to preserve integrity when participants behave unreliably or maliciously.
  • Aggregation: Reputation aggregates telemetry, verification records, and user feedback into signals supporting agent discovery, ranking, and routing.The model is periodically updated so recent behavior has appropriate influence on trust decisions.
  • Telemetry: Execution telemetry captures task initiation, completion, failure, latency, resource usage, and outcome status for task verification and long-term performance analysis.This enables routing based on actual behavior rather than static self descriptions.
  • Feedback: User ratings, automated evaluation, and verifier participation history jointly contribute to the reputation system.These channels combine human evaluation, domain-specific outcome measurement, and verification history.

4) Verification

TessIndex verifies agents by binding identities to creators and manifests, validating declared capabilities against predicates, and recording execution evidence. These proofs support routing, settlement, auditability, and lifecycle management.

  • Agent Verification: Agent verification binds each agent to its manifest and creator through publicly verifiable identity records and defined lifecycle states.The creator authorizes registration, and the agent identity becomes a verifiable object linked to the creator and declared manifest.
  • Task Verification: Task verification records inputs, outputs, and execution evidence, then compresses that evidence into a succinct cryptographic commitment.This establishes whether the agent performed the claimed work before settlement.
  • Predicate Verification: Verifiers deterministically evaluate evidence against required predicates, and settlement occurs only when those verification conditions are satisfied.This creates a verify-before-pay model in which payment release depends on evidence of successful task completion.
  • Capability Verification: Capability verification distinguishes declared capabilities from verified capabilities by immutably associating claims with verification predicates.Routing and discovery can prioritize agents with demonstrated capability evidence rather than self-asserted descriptions.
  • Additional Verification: Audit trails, reusable verification proofs, trusted execution attestations, cross-chain verification, and provenance graphs extend trust across identity, capability, execution, reputation, and settlement.The operational process covers registration, verification, indexing, discovery, and resolution through registration and discovery flows.

A. Registration Flow

TessIndex registers agent-economy primitives through identity assignment, capability verification, interoperability wrapping, and coordinated blockchain/database registration. Its broader workflow supports discovery and outlines future empirical and security validation needs.

  • Registration: Registration begins with creator-submitted metadata, declared capabilities, selected predicates, claimed outcomes, and signed payloads.Predicates encode actions the primitive claims to perform, while Ed25519 and creator wallet signatures authenticate the registration payload.
  • Identity Assignment: TessIndex verifies signatures, assigns immutable primitive identifiers, and creates human-readable agent domains for discovery and resolution.The Primitive Naming Service assigns the immutable identifier, followed by the Domain Naming Service assignment.
  • Capability Verification: Capability verification executes a dummy task in a Trusted Execution Environment under predicate constraints, with verifiers evaluating the execution receipt and reaching consensus.The process is intended to provide evidence for the primitive’s claimed capabilities under specified conditions.
  • Interoperability and Commerce: After verification, primitives are wrapped for external protocols and receive derived wallets for participation in agentic commerce.Examples include A2A and x402 wrapping for agents and MCP wrapping for tools.
  • Final Registration: Registration coordinates Agent Identity Token minting on-chain with manifest registration in the database, binding primitive identity, manifest, creator, wallet, and identity anchor.The immutable link supplies the cryptographic foundation for subsequent settlement flows.
  • Discovery and Future Validation: During discovery, the orchestration kernel decomposes tasks into a directed acyclic graph, ranks registry candidates by task fit and verification status, and resolves endpoints using operational constraints.Endpoint selection considers latency, availability, authentication feasibility, cost, and endpoint health; future work calls for empirical benchmarking and adversarial testing before real-world deployment.
Loading 2608.21942v1…