Create a practical OpenClaw integration package for AMCS so OpenClaw agents can use AMCS consistently as structured memory and project context without drifting into stale instructions or ad-hoc usage.
Why
This issue is still valid, but parts of the original idea are already present inside AMCS itself:
describe_tools exposes the live tool list
annotate_tool persists usage notes
project skills/guardrails already exist
llm/memory.md already documents a project startup flow
So this issue should now focus on the OpenClaw-side integration work, not on re-planning features AMCS already has.
Recommended scope
1. OpenClaw skill/package for AMCS usage
Create one or more OpenClaw-facing skills or instruction packages that teach agents to:
prefer AMCS when durable structured memory is appropriate
use project tools early in project-scoped work
load project context / skills / guardrails before implementation
use domain-specific AMCS tools when available instead of stuffing everything into generic notes
2. Practical startup pattern for AMCS-backed projects
Standardize an OpenClaw-side project startup pattern such as:
identify or set project
call describe_tools
load project context
load project skills
load project guardrails
then plan or implement
3. Storage-boundary guidance
Document when OpenClaw should prefer:
local workspace memory
AMCS thoughts
AMCS structured domain tools
AMCS skills/guardrails
4. Lightweight sync/check support
If needed, add a small helper or maintenance step for OpenClaw skill authors so AMCS guidance can be refreshed against the live AMCS tool list.
Out of scope
redesigning AMCS tool discovery from scratch
duplicating functionality that describe_tools and annotate_tool already provide
solving every integration surface beyond OpenClaw
Acceptance criteria
At least one OpenClaw-facing AMCS skill/package exists
There is a documented AMCS project startup flow for OpenClaw agents
Guidance clearly separates local memory vs AMCS thoughts vs structured AMCS tools
The integration uses the live AMCS tool surface instead of assuming a frozen tool list
The resulting guidance is small and maintainable
Notes
This issue now intentionally overlaps less with the old discovery/sync ticket, because the core runtime discovery problem is already largely addressed inside AMCS.
## Summary
Create a practical OpenClaw integration package for AMCS so OpenClaw agents can use AMCS consistently as structured memory and project context without drifting into stale instructions or ad-hoc usage.
## Why
This issue is still valid, but parts of the original idea are already present inside AMCS itself:
- `describe_tools` exposes the live tool list
- `annotate_tool` persists usage notes
- project skills/guardrails already exist
- `llm/memory.md` already documents a project startup flow
So this issue should now focus on the **OpenClaw-side integration work**, not on re-planning features AMCS already has.
## Recommended scope
### 1. OpenClaw skill/package for AMCS usage
Create one or more OpenClaw-facing skills or instruction packages that teach agents to:
- prefer AMCS when durable structured memory is appropriate
- use project tools early in project-scoped work
- load project context / skills / guardrails before implementation
- use domain-specific AMCS tools when available instead of stuffing everything into generic notes
### 2. Practical startup pattern for AMCS-backed projects
Standardize an OpenClaw-side project startup pattern such as:
1. identify or set project
2. call `describe_tools`
3. load project context
4. load project skills
5. load project guardrails
6. then plan or implement
### 3. Storage-boundary guidance
Document when OpenClaw should prefer:
- local workspace memory
- AMCS thoughts
- AMCS structured domain tools
- AMCS skills/guardrails
### 4. Lightweight sync/check support
If needed, add a small helper or maintenance step for OpenClaw skill authors so AMCS guidance can be refreshed against the live AMCS tool list.
## Out of scope
- redesigning AMCS tool discovery from scratch
- duplicating functionality that `describe_tools` and `annotate_tool` already provide
- solving every integration surface beyond OpenClaw
## Acceptance criteria
- [ ] At least one OpenClaw-facing AMCS skill/package exists
- [ ] There is a documented AMCS project startup flow for OpenClaw agents
- [ ] Guidance clearly separates local memory vs AMCS thoughts vs structured AMCS tools
- [ ] The integration uses the live AMCS tool surface instead of assuming a frozen tool list
- [ ] The resulting guidance is small and maintainable
## Notes
This issue now intentionally overlaps less with the old discovery/sync ticket, because the core runtime discovery problem is already largely addressed inside AMCS.
sam
changed title from Plan better OpenClaw integration for AMCS with skill-driven and tool-aware workflows to Create a practical OpenClaw integration package for AMCS2026-04-03 12:02:00 +00:00
Not sure if we need to do anything on openclaw. Much better projects out there.
Might just build a few usable skills and make recommendations for your AGENTS.MD
Not sure if we need to do anything on openclaw. Much better projects out there.
Might just build a few usable skills and make recommendations for your AGENTS.MD
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Summary
Create a practical OpenClaw integration package for AMCS so OpenClaw agents can use AMCS consistently as structured memory and project context without drifting into stale instructions or ad-hoc usage.
Why
This issue is still valid, but parts of the original idea are already present inside AMCS itself:
describe_toolsexposes the live tool listannotate_toolpersists usage notesllm/memory.mdalready documents a project startup flowSo this issue should now focus on the OpenClaw-side integration work, not on re-planning features AMCS already has.
Recommended scope
1. OpenClaw skill/package for AMCS usage
Create one or more OpenClaw-facing skills or instruction packages that teach agents to:
2. Practical startup pattern for AMCS-backed projects
Standardize an OpenClaw-side project startup pattern such as:
describe_tools3. Storage-boundary guidance
Document when OpenClaw should prefer:
4. Lightweight sync/check support
If needed, add a small helper or maintenance step for OpenClaw skill authors so AMCS guidance can be refreshed against the live AMCS tool list.
Out of scope
describe_toolsandannotate_toolalready provideAcceptance criteria
Notes
This issue now intentionally overlaps less with the old discovery/sync ticket, because the core runtime discovery problem is already largely addressed inside AMCS.
If the project being worked on can be identified, try to fetch/add it and set it as the active project.
Another requirement is that we add sessions on all the tools to be able to be compatible with Openclaw sessions etc.
Plan better OpenClaw integration for AMCS with skill-driven and tool-aware workflowsto Create a practical OpenClaw integration package for AMCSNot sure if we need to do anything on openclaw. Much better projects out there.
Might just build a few usable skills and make recommendations for your AGENTS.MD