You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Vijay and I are the founders of VantedgeAI (W22). We build FundOS, an operating system
for VC, PE and private credit funds covering compliance and operations for LP records,
capital calls, NAV, waterfalls, covenants and data rooms.
We have been testing QM against FundOS and have written seven skills that let an agent do
actual fund operations: draft a compliant capital call from the live LP roster, run a
distribution waterfall and check portfolio covenants.
We initially assumed we could point QM at our MCP server. It took reading
claude-harness.ts to understand that the harness uses MCP as its own internal
transport. That makes sense, but the SDK sitting in devDependencies sent us down the
wrong path. Maybe it makes sense to add a note in the README as it may save others the
same detour.
For us, the credential broker is the most valuable part. Some of these workflows can
affect real LP money, so having the agent never hold the credentials matters far more
than convenience. Maybe this feature should be highlighted more prominently.
We would be happy to PR the skills into skills-seed/ if they are useful. Otherwise we can
leave them in our repo. Let us know whether a vertical-specific fund operations pack makes
sense for the seed set.
Best,
Ravi
Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.
One more note in case it's useful for evaluating these.
The skills are plain REST over https://fundos.vantedgeai.com/api/v1 — no MCP, no SDK, no
build step. Each SKILL.md is readable on its own, so the quickest review is just opening fundos/SKILL.md (the root one: auth, the shell helper, the safety contract) and then any
domain skill next to it. Every example is a literal curl you can eyeball.
Two things we deliberately put in there that you may or may not agree with:
Credentials go through QM's broker rather than a token in the sandbox, so the agent never
holds the key and your admin pins host/method/path at QM's edge.
Writes are proposals. FundOS routes every write to a human approval queue server-side, and
the skills are written to report "drafted for approval", never "issued" — a capital call
moves real LP money and an agent should not be the last step.
If you want to actually run them rather than read them, tell me and I'll set you up with a
sandbox workspace and a credential — synthetic fund data, nothing real, so you can push it
around freely. Not posting one here for obvious reasons.
Happy either way — no need to run anything to tell us whether a vertical pack belongs in the
seed set.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Hi,
Vijay and I are the founders of VantedgeAI (W22). We build FundOS, an operating system
for VC, PE and private credit funds covering compliance and operations for LP records,
capital calls, NAV, waterfalls, covenants and data rooms.
We have been testing QM against FundOS and have written seven skills that let an agent do
actual fund operations: draft a compliant capital call from the live LP roster, run a
distribution waterfall and check portfolio covenants.
The skills can be found here:
https://github.com/8vdx1/fundos-mcp/tree/main/integrations/qm
Two points worth noting:
We initially assumed we could point QM at our MCP server. It took reading
claude-harness.ts to understand that the harness uses MCP as its own internal
transport. That makes sense, but the SDK sitting in devDependencies sent us down the
wrong path. Maybe it makes sense to add a note in the README as it may save others the
same detour.
For us, the credential broker is the most valuable part. Some of these workflows can
affect real LP money, so having the agent never hold the credentials matters far more
than convenience. Maybe this feature should be highlighted more prominently.
We would be happy to PR the skills into skills-seed/ if they are useful. Otherwise we can
leave them in our repo. Let us know whether a vertical-specific fund operations pack makes
sense for the seed set.
Best,
Ravi
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is disabled.