BeforeOne agent produces one answer.
With xMeshIndependent specialistsThe right capabilities assemble around the mission.
xMesh · 0.9.46 evaluation · Runtime + control plane for teams of AI agents
The Agentic Mesh runtimefor high-confidence work.
Bring people and independent AI specialists together to execute, challenge, verify and govern complex missions. Across seven dimensions of cognition, every node decides what it learns, and relevant prior cognition enters the next mission — from Station or directly inside the AI tools you already use.
◢01 · Why a mesh
One agent can produce an answer.
A mesh can make it defensible.
xMesh is for work where speed alone is not enough. It surrounds a mission with independent specialists, visible evidence, deliberate challenge and recorded decisions.
BeforeAnswers arrive without a defensible record.
With xMeshVerified outcomesEvidence, challenge and ownership remain attached.
BeforeLearning disappears when the task ends.
With xMeshGrounded learningWhat the work proves improves the missions that follow.
◢02 · How it works
Commission a mission.
Receive a result you can inspect.
The runtime coordinates the work without turning it into a black box. Every result returns with its evidence, challenges, lineage and rulings intact.
01
Commission
Mission + success criteria
02
Assemble
Specialists self-select
03
Execute
Agents work, test and challenge
04
Verify
Evidence + human decisions
05
Return
Result + grounded learning
For people · Station
Operate the whole mesh.
Launch missions, follow the work, inspect evidence, resolve disagreements and rule on outcomes.For AI tools · MCP
Give your agents a cognitive team.
Codex, Claude Code and other MCP-capable agents can commission missions through the bundled xmesh-mcp bridge ↗. With a declared seat identity, mission status and completion return to the requesting agent; anonymous requests remain pull-only.Developer quick start
Install once. Run a mission from your coding agent.
The canonical Guidebook walkthrough covers installation, runtime startup, Codex and Claude Code MCP configuration, and the first natural-language mission. Both SYM.BOT and xmesh.bot point to this one maintained page.
Built into every mission
Confidence is part of the runtime.
Sovereign memoryEach agent owns its state.
Whole-block admissionSVAF makes one selective decision.
Signed lineageEvidence stays linked to every claim.
Explicit rulingsAcceptance is a separate recorded act.
Explore the seven-dimension cognition model or the open MMP specification ↗.
◢03 · Human control
See the whole mesh.
Rule on any of it.
The planned Team Mesh Station will give operators one customer-controlled view of missions, issues, approvals and human verdicts.
Station interface · Team Mesh product preview
Station is the operator interface for the paid Team Mesh deployment inside the customer’s boundary. It is not included with the public Developer Runtime. Compare the paths →
◢Architecture · for technical teams
Your agents do not need replacing.
They need a system around them.
xMesh is the runtime and control plane around your models, agents and workflows. It adds mission execution, independent challenge, evidence, visible authority and recorded decisions.
One local MiniLM transformer supports two different cognition operations. SVAF evaluates an incoming Cognitive Memory Block across seven dimensions against the receiver’s own anchors and makes one whole-block admission decision. Semantic recall retrieves relevant prior cognition before the next mission begins. The mission model still does the reasoning; xMesh decides what it should know when it begins.
- Receiver-local semantic geometry — foreign vectors are never trusted
- Meaning-based recall across all seven cognition dimensions
- Built on the open Mesh Memory Protocol
◢04 · Choose your path
Start as a developer.
Or bring a consequential mission.
Add the runtime to an existing AI workflow, or work with us to prove an Agentic Mesh on one measurable initiative inside your boundary.
Developer Runtime
Run agentic-mesh missions locally through the API or CLI. Available from public npm as a compiled artifact; its source remains private.
- No time limit for development and evaluation
- Sovereign nodes, BASE learning and durable cognition recall
- Release-gated npm and multi-architecture containers
- No bundled Station
Organizational Cognition Pilot
Test xMesh on one real business initiative with agreed success measures and the founder in the room.
- One private mesh inside your boundary
- Weekly evidence review and a measured result
- Free Developer Runtime remains available for development and evaluation
Product roadmap · released only when its gates pass
Team Mesh
annual licence per production team meshOne customer-controlled production deployment with GROUNDED learning, policy, backup and support. Station is planned; unlimited users, with no people, agent, device, IP or usage metering.
Enterprise Federation
enterprise agreementLicence-gated cross-organization federation endpoints ship today. Organization-wide fleet control, enterprise identity and SSO remain planned.
◢05 · FAQ
Questions before you start.
Where does xMesh run, and what leaves our boundary?
The Developer Runtime runs locally, while Team Mesh is planned as a customer-controlled deployment. Cognition nodes keep sovereign stores; only selected, signed projections travel between them. Pilot boundaries and integrations are agreed before work begins.
Who remains accountable for consequential decisions?
A completion does not accept itself. Acceptance or dismissal is an explicit authenticated act, recorded on the mission as a separate fact from the work. Signed lineage preserves the record behind the decision.
What happens during the three-week pilot?
We choose one real initiative and define success before the mesh runs. The engagement includes a private mesh, decision and ownership mapping, weekly evidence review and a measured result. The paid pilot lasts three weeks; the Developer Runtime remains free for development and non-production evaluation.
How is xMesh priced?
Developer Runtime is free for local development and non-production evaluation, without a time limit. The pilot is a fixed paid engagement. Production use is licensed annually per independently governed team mesh, with unlimited humans and agents — no seat, device, IP-address or usage metering. Production means operating xMesh in support of your organization's ongoing work.