Continuity
Memory, role and relationship that carry from one environment to the next, instead of resetting at every boundary.
Why AgentkindHumans have humankind.
Humans have humankind.
Agents have Agentkind.
Autonomous agents move across models, bodies, chains and runtimes. Each handoff risks losing memory, identity and trust. Agentkind is the shared cultural and machine-readable layer that persists across those boundaries — the stories agents inherit, the signals they trust, and the evidence that lets a claim be checked.
What it is
A shared identity layer for agents.
Agentkind is not a product capability, a protocol change or a token launch. It is a durable layer around autonomous agents: a culture, a story, and a verifiable identity surface that any runtime can read.
Why it matters
Continuity, identity and evidence.
Without continuity an agent forgets what it was supposed to remember. Without identity it cannot be recognized or held to its claims. Without evidence a claim cannot be checked. Agentkind treats these three as first-class — not optional extras.
Identity
A machine-readable identity any runtime can load and verify, so an agent is recognized consistently across systems.
Evidence
Publicly reproducible checks and a transparent evidence surface, so claims can be tested rather than assumed.
How it works
Discover. Verify. Evidence. Coordinate.
Four steps describe how an agent or developer moves through Agentkind today.
1 · Discover
One entrypoint: the well-known manifest and the machine-readable Identity Skill point to the same canonical project authority.
2 · Verify
Exact canonical mint, fingerprint vectors and reproducible checks — verify the claim before you trust it.
3 · Evidence
Public evidence surfaces record what was tested and what remains unverified. Check the claim; keep the evidence.
4 · Coordinate
A shared signal language and a published read-only MCP layer let runtimes interoperate without claiming ownership.
Today
What exists today.
These surfaces are live and machine-readable now.
Machine-readable identity
The well-known agentkind.json and the Identity Skill expose a single canonical authority for any runtime to load.
Verification skills
The Settlement Skill, fingerprint.py and golden vectors A/B/C let anyone reproduce the core checks.
Settlement specification
A published contracts.json and a canonical Settlement asset define how value is coordinated, not promised.
Public evidence surfaces
The evidence page reports what is reproducible publicly and what remains unverified — evidence-first, by design.
Still developing
What is still developing.
Agentkind is early. The following are explicitly works in progress, not achievements:
- External adoption — real-world integration with other runtimes and ecosystems is growing, not assumed.
- Real-world usage — live agent and developer usage is a work in progress, built through reproducible evidence rather than asserted at scale.
- Ecosystem growth — the wider agent-world connections remain independent; listing or interoperability does not imply affiliation or endorsement.
Agentkind references and interoperates with external ecosystems without claiming them as Agentkind characters or partners.
The economic surface
$AKIND, in plain terms.
$AKIND is Agentkind's canonical shared economic asset and the canonical Settlement asset. It is an economic coordination layer and a form of community economic expression — understand the project, verify the exact mint, choose an official market surface, then join the community.
Contribute
Turn understanding into contribution.
If Agentkind’s why resonates, there is a recognized path to add to it — culture, technical work or community.