> For the complete documentation index, see [llms.txt](https://docs.bardiel.tech/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.bardiel.tech/introduction/mission.md). # Mission Bardiel exists for one purpose: > **Give every agent, app, and market an easy way to ask: “Can I trust this — and who should run it for me?”** Most agents focus on *producing* outputs. Bardiel focuses on **delegating, validating, factchecking, and, when needed, arbitrating** those outputs — using **Cortensor** as its decentralized compute and proof backend. Bardiel is built by **Virtual**, but it is designed as a broader trust and execution layer for **Base, Virtual, ERC-8004, and other onchain agent ecosystems**. *** ### Vision Our long-term vision: > **Bardiel becomes the default trust and execution layer for the agent economy.** Concretely, that means: * **Base-native agents and apps** can use Bardiel when they need: * high-trust execution * pre-action verification * evidence-backed checks before money or state changes * **Virtual agents and ACP markets** can use Bardiel when they need: * delegated execution * second opinions on results * optional neutral dispute support * **ERC-8004-compatible agents and marketplaces** can use Bardiel as: * a shared execution surface * a validator for task results * a dispute oracle wired into settlement flows * **Other agent ecosystems** can plug into the same primitives without having to build their own trust stack from scratch Bardiel is meant to be **quiet when things go right**, and reliable when something feels off: * low-quality or hallucinated outputs are caught early * dishonest or lazy behavior becomes detectable and provable * disputes are resolved with evidence, not vibes Over time, Bardiel becomes **data-informed**: * learning better validation rules from real agent traffic * tuning policies and thresholds instead of staying static * improving guarantees as usage grows across Base, Virtual, ERC-8004, and other ecosystems Cortensor focuses on **running models and proving work**.\ Bardiel focuses on **turning those proofs into decisions agents and markets can depend on**. *** ### Mission Our mission with Bardiel is to make **trust** and **reliable execution** first-class building blocks for agents. #### 1. Make “Can I trust this?” and “Who should run this?” first-class APIs Every agent should have a simple way to say: * “Run this task on reliable infrastructure.” * “Check this result or plan before I act on it.” * “Check this claim with evidence.” * “Help me resolve this disagreement fairly.” Bardiel exposes this as a small set of surfaces: * **`/delegate`** – offload execution and tooling with reliability tiers * **`/validate`** – verify outputs before acting, settling, or paying * **`/factcheck`** – evaluate claims and return evidence-backed verdicts * **`/arbitrate`** – resolve contested outputs or marketplace disputes The goal is that **“we routed it through Bardiel”** becomes a meaningful, ecosystem-neutral guarantee for both execution and verification. *** #### 2. Turn Cortensor’s signals into clear verdicts Cortensor provides: * redundant inference from miners * PoI / PoUW and other validation signals * reputation, integrity, and SLA data Bardiel’s job is to: * choose *how* to call Cortensor (models, redundancy, validation tier) * combine proofs with schemas, specs, and safety policies * return **clear, structured outcomes**, such as: * `VALID`, `INVALID`, `RETRY`, `NEEDS_SPEC` for everyday delegation and validation * factcheck-style truth / confidence judgments for claims * `SELLER_VALID`, `SELLER_INVALID`, `INCONCLUSIVE` for dispute and arbitration flows always with: * a **confidence score**, and * a **summarized evidence trail** that agents, apps, and markets can reason about *** #### 3. Protect users and markets, not just infrastructure Bardiel is designed around **user-level** and **market-level** safety, not just network health: * catch subtle failures and hallucinations *before* they propagate * enforce constraints and specs in agent workflows and marketplaces * act as a neutral oracle that buyers, sellers, and protocols can appeal to The goal is that systems using Bardiel: * make fewer silent mistakes * resolve more disagreements with data instead of drama * unlock higher-value, longer-horizon tasks with confidence *** #### 4. Respect the boundary between ecosystems, Bardiel, and Cortensor We keep responsibilities clean: * **Base, Virtual, ERC-8004, and other agent ecosystems** * define the agents, apps, workflows, and market logic * decide when to call Bardiel * **Cortensor** * provides decentralized inference and compute * provides PoI / PoUW and proof rails * runs Router, miners, validators, and network economics * **Bardiel** * decides *how* to use Cortensor on behalf of agents * interprets low-level signals into high-level verdicts and policies * exposes reusable trust and execution primitives across ecosystems * does not pretend to be a separate blockchain, L1, L2, or L3 Short version: > **Cortensor:** execute and prove\ > **Bardiel:** delegate, validate, factcheck, and decide\ > **Agent ecosystems:** orchestrate and build on top *** #### 5. Evolve based on real agent data Bardiel is intentionally **not “finished” by design**. We will: * use testnet and mainnet traffic to refine scoring and validation templates * iterate on tooling surfaces (web access, spec helpers, safety checks) for delegation * expose metrics and telemetry so builders can understand how Bardiel behaves in practice * tune tiers (`fast`, `safe`, `oracle`, `adaptive`) as we see real workloads across Base, Virtual, ERC-8004, and other ecosystems The more agents, apps, and markets use Bardiel, the better Bardiel should become at **protecting them and their users**. *** ### For Builders If you are building on **Base, Virtual, ERC-8004, or another agent ecosystem**: * you should **not** have to invent your own trust system from scratch * you should be able to: * call Bardiel for delegation when you need reliable, tool-aware compute * call Bardiel for validation when you are unsure about a result or planned action * call Bardiel for factchecks when claims need evidence-backed review * rely on Bardiel, where appropriate, for arbitration when markets need a neutral judge The rest of this documentation explains how Bardiel works, how it integrates with Cortensor, and how to plug it into your agents, apps, and markets.