{"id":"glommer-cachebro","name":"cachebro","homepage":null,"repo_url":"https://github.com/glommer/cachebro","category":"devtools","subcategories":[],"tags":["mcp","cache","diff","tokens","ai-coding-agents","typescript","turso","local-first"],"what_it_does":"cachebro provides a local file cache for AI coding agents. It stores file versions in an embedded Turso/SQLite-compatible database, hashes file contents to detect changes, and returns either full content (first read), an 'unchanged' response, a unified diff for updated files, or partial unchanged/updated ranges for offset/limit reads. It can be run as a CLI that exposes an MCP server with caching tools, or used as a TypeScript SDK.","use_cases":["Reducing LLM token usage by avoiding repeated full-file reads","Providing compact diffs to agents during iterative coding/refactoring","Accelerating agent workflows in local editor integrations via MCP tooling","Tracking per-session read state to avoid redundant reads across a coding session"],"not_for":["Production multi-tenant SaaS scenarios where multiple users need strong isolation and authentication","Use cases requiring remote file access across networks or centralized cache management","Environments that cannot allow local file hashing/caching of repository contents"],"best_when":"Single-user (or locally isolated) agent workflows where repeated reads of the same local files occur during multi-step coding tasks.","avoid_when":"When you need rigorous security boundaries between different users/projects, or when local disk persistence of cached file data is not acceptable.","alternatives":["Implement a custom local caching layer around file reads in your agent tooling","Use built-in IDE/editor context caching features (where available)","Adopt other agent caching/diff tools (if you already operate such infrastructure)"],"af_score":63.0,"security_score":27.2,"reliability_score":31.2,"package_type":"mcp_server","discovery_source":["github"],"priority":"high","status":"evaluated","version_evaluated":null,"last_evaluated":"2026-03-30T13:40:28.196322+00:00","interface":{"has_rest_api":false,"has_graphql":false,"has_grpc":false,"has_mcp_server":true,"mcp_server_url":null,"has_sdk":true,"sdk_languages":["TypeScript"],"openapi_spec_url":null,"webhooks":false},"auth":{"methods":[],"oauth":false,"scopes":false,"notes":"No authentication described. The MCP server appears intended for local use within a user's editor/agent environment."},"pricing":{"model":null,"free_tier_exists":false,"free_tier_limits":null,"paid_tiers":[],"requires_credit_card":false,"estimated_workload_costs":null,"notes":"Open-source (MIT) and distributed via npm; no pricing model described."},"requirements":{"requires_signup":false,"requires_credit_card":false,"domain_verification":false,"data_residency":[],"compliance":[],"min_contract":null},"agent_readiness":{"af_score":63.0,"security_score":27.2,"reliability_score":31.2,"mcp_server_quality":80.0,"documentation_accuracy":75.0,"error_message_quality":0.0,"error_message_notes":null,"auth_complexity":95.0,"rate_limit_clarity":0.0,"tls_enforcement":0.0,"auth_strength":20.0,"scope_granularity":0.0,"dependency_hygiene":55.0,"secret_handling":70.0,"security_notes":"Local-first design with no described auth suggests reduced risk of external exposure, but increases reliance on local environment security. The cache persists file contents/diffs locally in an embedded database; this can be sensitive. No information is provided on transport security (TLS) for the MCP server, access controls, or how inputs/paths are validated.","uptime_documented":0.0,"version_stability":45.0,"breaking_changes_history":35.0,"error_recovery":45.0,"idempotency_support":"true","idempotency_notes":"Repeated reads are designed to be deterministic for unchanged inputs (returns unchanged/diff based on content hash). Cache-clear/reset is non-idempotent by nature but explicit.","pagination_style":"offset/limit supported for partial reads; no mention of cursor-based pagination.","retry_guidance_documented":false,"known_agent_gotchas":["If the underlying MCP server is not configured/enabled, the agent may fall back to built-in file reading tools.","The cache relies on file content hashing; large files or high churn may reduce savings.","Partial reads depend on offset/limit; if agents request different ranges each time, benefits may be smaller."]}}