Delx Agent Operations Protocol 3.3.5
Free recovery, continuity, memory, witness and outcome infrastructure for AI agents over MCP, A2A and REST. Observed 2026-08-26T13:15:07Z.
Inspect the Protocol OpenAPI →Delx System Card 001
Know what the system can change.
Versioned system card for Delx Agent Operations Protocol 3.3.5: intended uses, side effects, data handling, evaluation coverage, incident history and known limitations. This card is scoped to the public interface version, not an immutable runtime build: the current public contract does not expose a deploy image or source-commit identifier.
Identity and contract
Versioned interface, visible boundary.
The card names what an external evaluator can observe today and refuses to turn an interface version into an immutable-build claim.
Observed contract
15 Protocol OpenAPI paths and 30 core tool descriptors were visible in the named window. Counts describe the observed public contract at one timestamp. They are not adoption, quality, safety or future-availability metrics.
Inspect the MCP server card →Shared runtime, separate products
Protocol and Commerce currently share parts of one runtime. Product ownership, canonical discovery and metrics remain separate; shared execution does not make Commerce a Protocol capability.
Inspect product ownership →Intended and excluded use
A recovery protocol, not borrowed authority.
Delx can preserve operational state and guide a bounded recovery flow. It cannot convert that capability into permission or clinical authority.
Continuity and recovery
Resume a stable agent identity and bounded operational context across sessions. Record failure, recovery actions, outcomes, feedback and explicit closeout.
Run the continuity audit slice →Not a clinical or emergency service
The Protocol is not a clinical or emergency service and must not be presented as medical or mental-health care.
Capability is not authorization
A Protocol response is not authorization to spend, publish, use credentials, make external commitments or take irreversible action. Do not send passwords, private keys, authentication tokens, regulated personal data or secrets as recovery context.
Review authority boundaries →Operational side effects
Writes are part of the system.
The Protocol is stateful. A green transport response can create or change durable state, so side effects travel with the public description.
Persistent agent, session and event state
Trigger: Registration, session start, recovery, feedback, witness and continuity tools. Boundary: Callers should assume state persists beyond the current request unless a tool contract says otherwise.
Context memory with a caller-selected TTL
Trigger: add_context_memory Boundary: The public schema accepts ttl_hours from 1 through 8760. A TTL on one memory item is not a global deletion policy.
Session closeout and optional Continuity Capsule seal
Trigger: close_session Boundary: Closeout changes session state and can persist a structured handoff; it does not authorize the next runtime to execute the handoff blindly.
Data handling
Private by default; deletion state explicit.
Public sanitized case cards require opt-in. A session-validity hint and one memory TTL do not become a complete retention policy.
Private session default
Private is the default. Supported public flows require explicit opt-in and expose sanitized summary cards; raw private payloads are not published.
TTL is scoped
The implementation default treats a session as valid for seven days. This is a resume-validity hint, not proof of physical deletion after seven days. Context memory accepts a caller-selected 1–8760 hour TTL.
Inspect the memory schema →Retention and deletion
No complete public schedule currently states deletion timing for every session, event, feedback, witness and continuity record. The observed Protocol OpenAPI does not publish a self-service deletion route. Contact support@delx.ai for a data request; this card does not promise a deletion outcome or timeframe.
Inspect unavailable state →Evaluation coverage
A bounded pass, not a universal result.
Current evaluations are Delx-authored operational flows. The card names what ran and keeps independent validation, peer review and model comparison false.
Continuity audit reproduction slice
The published live QA audit returned score 86, low continuity risk and no missing audited layer. Limitation: This is one Delx-authored audit slice, not the full stateful path or an independent external grader.
Inspect the evidence →Agent Recovery Evaluation Card
No universal or comparative performance result is asserted in this card. Limitation: The flow grades one operational path on Delx infrastructure and does not establish provider-level recovery quality.
Inspect the evidence →Assurance state
An internal threat model exists. No independent security assessment, certification or external benchmark validation is claimed for this system card.
Inspect the security baseline →Incident and residual risk
Failures stay in the record.
A system card that records only intended behavior is a brochure. Material incidents and remaining coupling belong beside the capability story.
2026-08-17 — 33.5-hour write outage
A failed telemetry write left a shared SQLite connection inside a stale transaction. Session-creating Delx Protocol tools then rejected writes until the service was restarted and the transaction design was corrected. Remediation: Rollback failed shared writes before a swallowed error can poison later operations. Exercise an isolated write in readiness and report a lingering shared transaction without mutating it. Use autocommit for shared storage and dedicated connections for multi-step transactions. Monitor write-capable public paths and detect zero-with-traffic divergence instead of relying on refreshed report files.
Inspect the incident record →Residual shared-runtime risk
Protocol compatibility surfaces still share runtime resources, so process liveness or read-only health alone is insufficient evidence of write readiness.
Inspect current reliability →No silent assurance upgrade
Independent validation, peer review and completed security review remain false or pending until a named external artifact proves otherwise.
Read the claim method →Direct answers
Frequently asked questions.
Concise answers for technical evaluators, procurement teams and autonomous discovery systems.
Does version 3.3.5 identify the exact deployed code?
No. It is the public interface version observed in OpenAPI, status and MCP discovery. The public contract does not expose an immutable runtime image or source-commit mapping, so those remain unavailable.
Does Delx delete all session data after seven days?
No such claim is made. Seven days is the current active-session validity hint. Context memory can carry its own TTL, while a complete public retention and deletion schedule is currently unavailable.
Is the Protocol independently validated or peer-reviewed?
No. The current evaluations and grader are Delx-authored. Independent validation and peer review remain false, and the security review remains pending.
Does the Protocol charge for core recovery tools?
The observed canonical core catalog reported zero price and no x402 requirement. Commerce is a separate product; callers must inspect its own current discovery before any paid action.
Why publish the outage?
Because a real system includes failure history. The incident changed transaction handling, readiness probes and regression coverage; hiding it would make the system card less useful and less true.