A command centre for agent orchestration, a native Salesforce data dictionary, and a delta-based CI/CD blueprint and tooling.
These aren't three separate projects so much as one connected system. Fulltrace is the factory that produces the rest: it's wired in two halves, a context layer (memory and skills over MCP) that Claude, Codex and Cursor plug into, and a runtime that runs workflows and audits, reaching models through OpenRouter. The halves are the machinery; everything in green is what comes off the line. That runtime builds and maintains all three projects and this site, and it audits Fulltrace itself, writing what it learns back into the shared memory. The two Salesforce tools sit either side of one source of truth: the CI/CD pipeline and tooling deploy metadata into the repo and org, and the data dictionary reads it back out as living documentation, with Confluence feeding documentation in through the tooling.
Purple is the model side: Claude, Codex and Cursor, plus the OpenRouter gateway.
Amber is Fulltrace, split into its context layer and its workflow runtime, which loops back to audit itself and top up that memory.
Green is what comes off the line: this site and the two Salesforce tools. The two tools work against one shared source of truth in the middle: the tooling deploys into it, the dictionary reads it back out. Confluence feeds documentation in through the tooling.
flowchart TD
CL["Claude"] --> AOSC
CX["Codex"] --> AOSC
CR["Cursor"] --> AOSC
OR["OpenRouter"] --> AOSR
subgraph AOS["Fulltrace"]
AOSC["Context layer
memory & skills over MCP"]
AOSR["Runtime
workflows & audits"]
end
AOSC --> AOSR
AOSR -->|"updates memory"| AOSC
AOSR --> WEB["This website"]
AOSR --> CI["CI/CD Pipeline
+ VSCode tooling"]
CI -->|"deploys"| SRC[("Git repo +
Salesforce metadata")]
SRC -->|"living docs"| DD["Salesforce Data Dictionary"]
CONF["Confluence"] --> CI
%% Blueprint tokens as literal hex (Mermaid cannot parse CSS vars or color-mix)
classDef model fill:#8b93e81f,stroke:#8b93e8,stroke-width:1.5px;
classDef entry fill:#e0be621f,stroke:#e0be62,stroke-width:1.5px;
classDef code fill:#62d99a1f,stroke:#62d99a,stroke-width:1.5px;
classDef docs fill:#8a8f9b14,stroke:#8a8f9b,stroke-width:1.2px,stroke-dasharray:4 3;
class CL,CX,CR,OR model;
class AOSC,AOSR entry;
class WEB,DD,CI code;
class CONF docs;
%% Fulltrace grouping box: dashed amber outline, no fill, matching the amber nodes it holds
style AOS fill:transparent,stroke:#e0be62,stroke-width:1.5px,stroke-dasharray:5 4;
One shared memory makes my AI assistants genuinely useful: a context portfolio served over MCP, plus shared skills and memory across Claude, Codex and Cursor, tied together by a VSCode dashboard. They know my work, act on their own, and everything they change is checked before it lands. It helps maintain this website.
A native, repository-based data dictionary with seven LWC viewers and async caching. It reads the blueprint Salesforce already keeps of your setup and turns your repo into searchable, always-current documentation, right inside Salesforce.
A vendor-agnostic release pipeline plus two editor extensions: delta-based configs for 5 CI platforms, guided setup, safe daily deployments, and automatic rollback when something fails.