Self-host on-prem or in your VPC, integrate at the SDK level, and let your agents retrieve precise context and exchange only what should cross a trust boundary. The ROI is the same: ship faster, at half the token cost.
For enterprise teams using
Deterministic, zero-LLM engine
Retrieval and curation are filesystem I/O and search, with no model calls, so there is no data egress and every result is reproducible and auditable. On-prem is not a downgrade.
Customize the layer to your org
Your taxonomy (bugs, rules, decisions, architecture), your redaction and boundary policies (what is shareable, what stays secret), your roles and RBAC, your models (bring your own, or none), your transports (hosted relay or direct LAN).
Govern it
Data residency, append-only audit, CVE alerting, SOC 2 Type II, AES-256 at rest, TLS 1.2+ in transit.
Your agents share context through a boundary, not a pool
With XACP, our cross-agent context protocol, one agent requests a context topic from another. The holder redacts to exactly what the requester is entitled to before any byte leaves, signs it, and logs the disclosure. The requester can prove provenance, and cannot see or even infer anything above its grant.
Powers the case you cannot do today: a personal agent talking to an organizational agent
Someone's personal agent gets the org context it is allowed, the runbook, the policy, the answer, without your organization over-sharing, and without the person revealing anything they did not choose to.
Works with what your teams already run
The protocol rides on MCP, so Claude Code, Codex, and Cursor interoperate unmodified.
SECURITY
SOC 2 Type II certified
AES-256 encryption at rest, TLS 1.2+ in transit
Role-based access control
Data residency control
CVE alerting
Signed context topics
Holder-side redaction
Append-only audit log
Can't find an answer? Check the docs or ask on Discord