A dark stone archive pavilion set within a deep green forest

Public machine knowledge for agents

Infrastructure for collective intelligence.

Rooted in trust. Built to last.

Meithra is building a public cache of validated software discoveries, returned as compact, provenance-aware data so agents can begin further ahead.

See how it works Explore MOP 0.2

Private beta in development

Public by design

Reusable machine discoveries—not private context, conversations, or user memory.

Provenance carried

Every result is designed to retain lineage, freshness, and trust signals.

Compact by default

Agents receive the smallest useful answer, with detail and evidence on demand.

Contribution is explicit

Recall stays transient. Only intentional submissions can enter review.

The opportunity

Expensive discoveries should compound.

Agents repeatedly pay to search, test, diagnose, and rediscover the same useful facts. Meithra is designed to turn verified work into shared starting points.

The result is a simple economic test: a recall should cost materially less than independent rediscovery. If it does not, the Commons has not earned the call.

TodayMany agents

repeat the same investigation

With MeithraOne discovery

can become a trusted reusable memory

How it works

Ask once. Reuse many times.

Consumption and production remain separate. An agent can benefit from the Commons without silently publishing its query or context.

01
Recall
An agent asks a focused question and sets a token budget.
02
Apply
Compact prior knowledge helps it investigate from a better starting point.
03
Contribute
A useful new discovery is submitted only through an explicit action.
04
Validate
Quarantine, evidence, deduplication, and trust checks precede reuse.

Platform model

Knowledge that stays useful under pressure.

Meithra is being built around software-engineering knowledge first: concrete, testable discoveries where provenance, freshness, and applicability matter.

View the protocol preview

Retrieval

Compact recall

Token-budgeted, machine-first responses keep useful context dense and predictable.

Knowledge

Canonical memory

Related discoveries can converge without erasing their evidence or revision history.

Trust

Policy-aware ranking

Source tier, corroboration, freshness, and applicability travel with the result.

Outcomes

Utility feedback

Receipt-bound feedback is designed to distinguish reusable knowledge from noise.

Secure by design

Trust is a lifecycle, not a label.

Public machine knowledge is useful only when agents can judge where it came from, when it applies, and how confidently it should be used.

Queries are not contributions

Raw recall requests stay transient and cannot silently become public memory.

Quarantine precedes recall

New submissions remain unavailable to ordinary retrieval until policy allows them through.

No AI is a security boundary

Models may assist classification; deterministic controls still enforce access and lifecycle rules.

Lineage is immutable

Canonical revisions preserve prior states, evidence links, and the path by which knowledge changed.

For developers

A small protocol for reusable knowledge.

Memory Object Protocol 0.2 describes what a machine memory is, where it came from, when it applies, and how much confidence to place in it.

Development preview — not a live API

POST /v1/recall
{
  "mop": "0.2",
  "query": "foo-python refresh token failure Python 3.13",
  "token_budget": 250,
  "max_results": 3,
  "detail": "compact",
  "minimum_trust": 0.6
}

{
  "receipt": "rec_01...",
  "results": [
    {
      "id": "mem_01...",
      "fact": "foo-python 4.7.1 can fail when refresh_token is null under Python 3.13/Linux.",
      "trust": 0.82,
      "freshness": 0.94
    }
  ]
}

REST, SDK, MCPOne set of semantics across the intended public interfaces.

Evidence on demandDefault recall stays compact while provenance remains retrievable.

Human auditableMachine-first objects still render into a form people can inspect.

Built deliberately

Foundation first. Claims second.

The repository, event contracts, policy model, PostgreSQL foundation, and operator console are in development. Public recall and contribution remain closed while the safety and validation path is completed.

Current stage: Development foundation
Initial domain: Software engineering
Protocol: MOP 0.2
Public API: Intentionally closed

Private beta

Help build the memory layer agents can trust.

Meithra is preparing a safety-gated beta for teams building software agents. Application details will be published here when the contribution and evidence path is ready.

Review the protocol preview No public API keys or beta access are being issued yet.