Loading
mcpcall is built by a consulting developer for consulting developers. It solves one specific problem: keeping years of personal AI-context work — CLAUDE.md, subagents, skills, slash commands, naming conventions, per-client tweaks — out of repos owned by clients.
The standard workaround is manual. Copy markdown into the project. Add the names to .gitignore. Hope no one runs git add -A. The shared .gitignore is committed, so even the file names leak. And there is no clean off-boarding when the engagement ends.
mcpcall replaces the workaround with infrastructure: a SaaS platform that stores your context files per workspace, and a stdio MCP server that pulls them into any local checkout. Files land in .claude/ and get hidden via .git/info/exclude — the per-repo, user-local ignore file that is never tracked and never pushed. The client sees neither the files nor the ignore rule. When the work ends, one cleanup call leaves the repo bit-identical to before.
Independent consultants and contracting agencies who work across many client repos and have invested in their own AI tooling. If you have a personal context library you do not want to give away with every engagement, this is built for you.
No team workspaces yet — single-owner only. No billing. No public signup. No team permissions, no SSO, no branching on file versions. All real refactors, all on the roadmap. None of them block the closed-beta loop.
Free beta. Signup is open — magic-link email or GitHub OAuth, your pick. No card on file, no quotas yet. The MCP server ships on npm as @asad120414/mcpcall — npx -y @asad120414/mcpcall is enough; no global install. The MCP server, the data model, the migrations, and this app live in one monorepo on GitHub. Issues and PRs welcome.