STANDARD · Profile revision 1

Model Context Protocol

A JSON-RPC protocol for connecting AI applications to tools, resources, prompts, and related capabilities, with W3C trace-context propagation.

standardmcptoolsjson-rpctrace-context

Why it is in the map

MCP is a high-signal agent-tool protocol with structured calls/results and trace propagation, while its limits expose the difference between tool response and external effect.

Tracevity boundary

A documented field can support reconstruction, but its presence alone does not prove completeness, authenticity, authorization, or external settlement.

Profile findings

What the documentation can—and cannot—establish

Each row distinguishes source evidence from Tracevity's bounded interpretation. “Not documented” describes the reviewed sources; it does not prove an implementation cannot record the artifact.

01

System identity

Partially documented

Protocol endpoint metadata and request/trace correlation are documented.

Tracevity interpretation

Request correlation is strong, but client/server metadata is self-reported and does not authenticate an agent or principal.

Limit

No complete agent definition, runtime instance, session, or subagent identity model is defined.

Primary evidence (1)
02

Principal and delegation

Partially documented

Authorization can constrain protocol access, but a complete principal/delegation trace artifact is not defined.

Tracevity interpretation

Authorization at runtime may exist without preserving initiating principal, delegated identity, credential scope, expiry, or elevation as evidence.

Limit

Identity-provider and authorization logs require separate correlation.

Primary evidence (1)
03

Instruction and context

Partially documented

Protocol-level contextual inputs and referenced resources are representable.

Tracevity interpretation

MCP does not provide a complete trace of the surrounding agent's user/system instructions or all selected context.

Limit

Context fingerprints, version guarantees, and explicit missing-context semantics are not uniformly defined.

Primary evidence (1)
04

Decision artifacts

Not documented

Agent decision artifacts are outside the core MCP trace semantics.

Tracevity interpretation

Tool selection rationale and hidden reasoning must remain unknown.

Limit

Applications may add metadata, but it is not standardized decision evidence.

Primary evidence (1)
05

Model activity

Not documented

Model activity tracing is outside the core protocol.

Tracevity interpretation

A sampling request is not equivalent to a complete model trace.

Limit

Model telemetry requires another semantic convention or runtime trace source.

Primary evidence (1)
06

Tool and MCP activity

Partially documented

Core MCP tool identity, request, arguments, result, and error semantics are documented.

Tracevity interpretation

The protocol reconstructs calls/results; complete timing, retries, server identity, and downstream span correlation depend on telemetry and implementation.

Limit

Related OpenTelemetry MCP spans remain Development status.

Primary evidence (1)
07

Effects

Not documented

External effect evidence is not standardized beyond tool request and reported result.

Tracevity interpretation

A successful MCP result cannot establish that an external destination changed state.

Limit

Destination evidence must be correlated separately.

Primary evidence (1)
08

Human control

Partially documented

Human confirmation is recommended, while complete operational control evidence is not standardized.

Tracevity interpretation

A design recommendation is not proof that approval, rejection, edit, interruption, takeover, or rollback request was recorded.

Limit

Approval chronology and exact event fields remain implementation-specific.

Primary evidence (1)
09

Outcome

Partially documented

Protocol and tool-reported outcome states are documented.

Tracevity interpretation

These states do not independently verify external success, rollback, compensation, or resulting state.

Limit

Partial and ambiguous business outcomes require application or destination evidence.

Primary evidence (1)
10

Trace integrity

Not documented

Protocol correlation is documented; trace integrity is not.

Tracevity interpretation

Self-reported endpoint identity and W3C trace context are not evidence-custody guarantees.

Limit

Runtime, collector, and backend controls require separate assessment.

Primary evidence (1)
11

Portability

Partially documented

Protocol interoperability and trace-context propagation are documented.

Tracevity interpretation

Telemetry portability remains split across MCP and a Development-status OpenTelemetry semantic layer.

Limit

No general vendor export, import, conversion, or semantic-loss matrix is defined.

Primary evidence (1)
12

Privacy

Partially documented

Protocol-level privacy and security cautions are documented.

Tracevity interpretation

These safeguards do not determine what telemetry captures or how a trace backend stores it.

Limit

Trace retention, deletion, redaction, hashing, code capture, and customer-controlled archival are not defined.

Primary evidence (1)

Reconstruction reading

Do not collapse these findings into one score.

This profile describes documented evidence surfaces. Whether they are sufficient depends on the reconstruction question and the other identity, authorization, tool, and destination records available.

Potentially useful evidence

  • Identity: Protocol endpoint metadata and request/trace correlation are documented.
  • Principal: Authorization can constrain protocol access, but a complete principal/delegation trace artifact is not defined.
  • Context: Protocol-level contextual inputs and referenced resources are representable.
  • Tools: Core MCP tool identity, request, arguments, result, and error semantics are documented.
  • Human control: Human confirmation is recommended, while complete operational control evidence is not standardized.

Explicit gaps or uncertainty

  • Decisions: Applications may add metadata, but it is not standardized decision evidence.
  • Models: Model telemetry requires another semantic convention or runtime trace source.
  • Effects: Destination evidence must be correlated separately.
  • Integrity: Runtime, collector, and backend controls require separate assessment.

Source register

5 reviewed primary sources

  1. Overview - Model Context ProtocolModel Context Protocol project · STANDARD SPECIFICATION · observed August 28, 2026 · current
    Open source ↗
  2. Tools - Model Context ProtocolModel Context Protocol project · STANDARD SPECIFICATION · observed August 28, 2026 · current
    Open source ↗
  3. Logging - Model Context ProtocolModel Context Protocol project · STANDARD SPECIFICATION · observed August 28, 2026 · current
    Open source ↗
  4. MCP 2026-07-28 changelogModel Context Protocol project · OFFICIAL RELEASE NOTES · observed August 28, 2026 · current
    Open source ↗
  5. OpenTelemetry MCP span modelOpenTelemetry project / CNCF · STANDARD SPECIFICATION · observed August 28, 2026 · current
    Open source ↗

Next question

What evidence does your use case require?

Move from documented field presence to an explicit reconstruction target, or examine selected directional format mappings without changing this system's documentation posture.

Trace Reconstruction Requirements Explore compatibility evidence