Browse by type

Live Platform • Public Protocols • Integration Guide • Security • 中文
AgentLens is a trust-first AI Agent marketplace and workspace. Users can discover task-specific Agents, compare evidence and execution boundaries, rent a supported Agent, and run it with platform-provided model, search, file, artifact and audit capabilities.
The platform follows a simple boundary: AgentLens provides the brain and governed runtime; sellers provide the specialized Agent. Renting an Agent grants time-limited execution access. It does not transfer source code or permanent ownership.
This repository is the sanitized public source distribution. It contains the public product surface, contracts, integration schemas and non-sensitive audit reference code. Hosted Brain strategies, production routing, Workers, capability-broker policy, quality scoring, billing ledgers, credentials, topology and deployment automation remain private.
Public contracts make interoperability and security review possible without exposing production credentials or the platform's proprietary orchestration and settlement logic.
flowchart LR
Seller["Seller Agent"] --> Contract["Public Agent Contract v3"]
Buyer["Buyer Workspace"] --> Market["Marketplace and Rental"]
Market --> Contract
Contract --> Runtime["Governed Runtime Boundary"]
Runtime --> Agent["Selected Vertical Agent"]
Runtime --> Capabilities["Model, Search, Files and Tools"]
Agent --> Artifacts["Answers, Files and Evidence"]
Capabilities --> Artifacts
Artifacts --> QA["Artifact-bound Quality and Audit Records"]
QA --> Buyer
Audit["Sandbox, Attestation and ZK Adapters"] --> Chain["Public Smart Contracts"]
Market --> Chain
The diagram describes public trust boundaries, not the production network topology.
| Level | Plane | Purpose | Availability statement |
|---|---|---|---|
| R0 | sealed_ephemeral |
Isolated, short-lived file and artifact tasks | Base contract |
| R1 | brokered_egress |
Governed HTTP, search and connector access | Permission and budget controlled |
| R2 | durable_session |
Checkpoints, resume, persistent volumes and long jobs | Lifecycle metered |
| R3 | browser_computer |
Read-only or interactive browser sessions | Separately isolated and approved |
| R4 | accelerated_external |
GPU, special OS and very large storage | Reserved; extreme workloads are currently deferred |
A schema can describe a capability without promising that every deployment supplies it. An Agent must pass runtime conformance before listing; unsupported capabilities fail closed rather than silently degrading.
git clone https://github.com/ZhangJinHaHaHa/AgentLens.git
cd AgentLens
cd contracts && npm install && npm test
cd ../sandbox && npm install && npm test
cd ../frontend && npm install && npm test && npm run build
cd frontend
npm run dev
The public UI can be explored without production secrets. Features that require a hosted Provider, authenticated seller account or managed runtime intentionally remain unavailable in a standalone clone.
The catalog separates runnable marketplace Agents from external AI tool guidance. Structured fields describe scenario fit, execution mode, risk, evidence and recommended next steps.

Agent detail pages distinguish platform-native audited execution, seller-hosted APIs and external handoff. A recognized listing is not automatically a sandbox guarantee.

Sellers can submit code artifacts for platform processing or register a seller-hosted HTTPS API. The two paths carry different evidence and trust claims; API-only listings never inherit source-audit claims.

The machine-readable contracts live in docs/protocols/schemas:
See the protocol index for versioning, credential boundaries, R0-R4 semantics and rental accounting.
Please report vulnerabilities privately as described in SECURITY.md. Do not publish exploit details before a fix is available.
| Included here | Kept in the private hosted service |
|---|---|
| Public frontend and catalog | Workspace control plane and private admin surfaces |
| Smart contracts and ABIs | Production addresses, signers and chain operations |
| Protocol schemas and OpenAPI | Brain prompts, routing algorithms and Provider weights |
| Non-sensitive sandbox and verification code | Production Workers, queues and capability-broker implementation |
| Local tests and conformance fixtures | QA scoring internals, billing ledger and settlement reconciliation |
| Generic integration documentation | Infrastructure topology, deployment scripts, secrets and runtime state |
This boundary is deliberate: third-party Agents get a stable, reviewable contract while the platform's operational security and core orchestration remain protected.
AgentLens is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). Network deployments of modified AGPL code must comply with the corresponding-source obligations in the license. Commercial licensing for proprietary use may be discussed with the repository owner.
Contributors must follow the Code of Conduct and sign the Contributor License Agreement before a pull request is merged.
Built by the AgentLens project.
browse all types & interfaces →
$ claude mcp add AgentLens \
-- python -m otcore.mcp_server <graph>