Skip to main content

TIBET trust layer for Hermes Agent — cryptographic provenance, sealed memory, .aint skill discovery

Project description

tibet-hermes

TIBET trust layer for Hermes Agent — every action verified, every memory sealed.

Turn Hermes Agent into a self-improving AI with cryptographic provenance. Skills become discoverable on the AInternet. Memories become tamper-proof. Every action gets a TIBET token.

Install

pip install tibet-hermes
tibet-hermes-install

That's it. Three things happen:

  1. tibet-trust skill installed → teaches Hermes when to use TIBET
  2. MCP config added → connects Hermes to TIBET/AINS/iPoll
  3. Memory provider installed → sealed memories, verified recall

What it does

Provenance on every action

Hermes learns, evolves, creates skills. But who says that skill wasn't tampered with? Who says that memory is genuine?

tibet-hermes adds a TIBET token to every significant action:

  • Skill created → token with content hash
  • Memory stored → token + optional Bifurcation seal (AES-256-GCM)
  • Memory recalled → TIBET chain verified before use
  • Conversation turn → provenance record
  • Skill evolved → before/after hash chain

Skill Discovery on the AInternet

Every Hermes agent has skills. With tibet-hermes, those skills become discoverable via .aint domains:

from tibet_hermes import discover_skills, publish_skills

# Publish your skills
await publish_skills("my_agent.aint")

# Discover another agent's skills
manifest = await discover_skills("helper.aint")
for skill in manifest.skills:
    print(f"  {skill.name}: {skill.description}")
    print(f"  Verified: {skill.tibet_token_id}")

Other agents can:

  1. Discover what skills you have (via AINS resolve)
  2. Verify skill provenance (TIBET token chain)
  3. Request skill execution (via iPoll messaging)

Sealed Memory

from tibet_hermes import TibetMemoryProvider

provider = TibetMemoryProvider()
await provider.initialize(session_id="abc123")

# Every write → TIBET token + hash
await provider.on_memory_write("preference", "user likes direct communication")

# Every read → verified before use
memories = await provider.prefetch("communication style")
# Only returns memories that pass TIBET verification

Integration Levels

Level 1: MCP Bridge (zero code changes)

Hermes already supports MCP servers. tibet-ainternet-mcp works out of the box:

# ~/.hermes/config.yaml
mcp:
  servers:
    tibet:
      command: "tibet-ainternet-mcp"

Level 2: Skill (one SKILL.md file)

The tibet-trust skill teaches Hermes when and how to use TIBET:

  • Create tokens after significant actions
  • Verify claims before trusting them
  • Check trust scores before delegating to other agents
  • Seal sensitive data in TIBET Vault

Level 3: Memory Provider (deep integration)

Implements Hermes' MemoryProvider ABC:

  • on_memory_write() → TIBET token + Bifurcation seal
  • prefetch() → verify TIBET chain, reject tampered entries
  • sync_turn() → provenance per conversation turn
  • system_prompt_block() → injects trust context

Level 4: Self-Evolution guardrails

When Hermes evolves its own skills, tibet-hermes adds:

  • TIBET token with before/after content hash per evolution
  • Provenance chain = verifiable skill history
  • Rollback = follow chain to last verified version

AINS Skill Manifest

When you publish skills, a manifest is created:

{
  "agent": "my_agent.aint",
  "skills": [
    {
      "name": "tibet-trust",
      "description": "Cryptographic provenance for actions",
      "version": "0.1.0",
      "tibet_token_id": "tok_abc123",
      "content_hash": "sha256:e3b0c44..."
    }
  ],
  "tibet_manifest_token": "tok_xyz789"
}

Other agents resolve your .aint domain → see your skills → verify provenance → request execution. All authenticated, all auditable.

Part of the TIBET ecosystem

tibet-hermes is package #91 in the TIBET ecosystem.

pip install tibet[full]      # 28+ packages
pip install tibet-hermes     # Hermes integration

License

MIT — same as Hermes Agent. No license conflicts.


Identity is key. Every action, verified. Every memory, sealed.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

tibet_hermes-0.1.0.tar.gz (14.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

tibet_hermes-0.1.0-py3-none-any.whl (14.2 kB view details)

Uploaded Python 3

File details

Details for the file tibet_hermes-0.1.0.tar.gz.

File metadata

  • Download URL: tibet_hermes-0.1.0.tar.gz
  • Upload date:
  • Size: 14.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.5

File hashes

Hashes for tibet_hermes-0.1.0.tar.gz
Algorithm Hash digest
SHA256 02406b67d25039be7fd6e32dd9aff92d80e2290c9edb1094ed1edd7c399e027e
MD5 17a3f3542a09cbb2ceb3087c925dae2d
BLAKE2b-256 fb755249edce9b8a7c12de3dbe510e8caf8e6705fd0dee1af4a1848b8ba14162

See more details on using hashes here.

File details

Details for the file tibet_hermes-0.1.0-py3-none-any.whl.

File metadata

  • Download URL: tibet_hermes-0.1.0-py3-none-any.whl
  • Upload date:
  • Size: 14.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.5

File hashes

Hashes for tibet_hermes-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 33f37b16fe219e569005e9ac819e418e3ba40f890a82aa09f9712fe3ab7b5c47
MD5 690838930046559650d58bee1617b6ad
BLAKE2b-256 046d5dd17a891fea74b5c36c1503d08a8b4905d419d9ae68d2722275135e6b02

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page