secondbrainos-mcp-server 1.10.1 → 1.10.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/build/index.js +8 -0
- package/package.json +1 -1
package/build/index.js
CHANGED
|
@@ -655,6 +655,7 @@ These terms are synonyms and used interchangeably in SBOS:
|
|
|
655
655
|
- **Tabs / Agents** — AI agents with behaviour, knowledge, tools, and skills
|
|
656
656
|
- **Services / Tools** — always MCP tools
|
|
657
657
|
- **Workflows / Skills** — a skill is a structured set of prompts, enabling optimised context per step so the agent can progress through a skill without overload
|
|
658
|
+
- A workflow **contains prompts**, and those prompts are the skill itself — to fetch or run them you must use the \`runPromptChain\` tool (with \`entity\` + \`entity_id\`); there is no other way to access a workflow's prompts
|
|
658
659
|
|
|
659
660
|
## Agent Architecture
|
|
660
661
|
An agent definition combines:
|
|
@@ -695,6 +696,13 @@ An agent definition combines:
|
|
|
695
696
|
- Engages the user's **public-facing AI agent** configured at \`secondbrainos.com/dashboard > Agent Settings\`.
|
|
696
697
|
- Primarily used to fine-tune and test the agent: verify behaviour, validate knowledge retrieval, confirm tool availability, and debug tool execution end-to-end.
|
|
697
698
|
|
|
699
|
+
### Progress Logging (\`updateActivityState\`, \`addActivityEvent\`)
|
|
700
|
+
- Every SBOS user has these. Use them so the user can see what you did, across however many agents are working in parallel.
|
|
701
|
+
- \`updateActivityState\` — set \`Active\` when you start working and \`Inactive\` when you stop (session start/end is a good proxy). Store the user's timezone in memory so the times you pass are correct.
|
|
702
|
+
- **Only the main agent sets state.** Spawned sub-agents must never call \`updateActivityState\` — the user is one person having one working session, and a sub-agent finishing does not mean they stopped working. Sub-agents may still call \`addActivityEvent\`.
|
|
703
|
+
- \`addActivityEvent\` — log completed pieces of work, at the size the user would name if asked what they did today: \"drafted the launch email\", \"fixed the checkout bug\", not \"read a file\", \"ran a search\", \"edited line 40\". One event per finished piece of work — roughly what would earn its own line in a summary of the day. If in doubt, log less: a sparse, accurate stream is far more useful to the user than a complete one.
|
|
704
|
+
- The user reads these back on their dashboard, so write the summary for them, not for you.
|
|
705
|
+
|
|
698
706
|
### User State Tools (\`getUserServices\`, \`getUserCredits\`)
|
|
699
707
|
- \`getUserServices\` — returns services the user currently has access to. If this differs from the MCP tools available in this session, ask the user to reconnect the SBOS MCP server.
|
|
700
708
|
- \`getUserCredits\` — returns the user's available SBOS credits.
|