All features, one platform
From package docs to private repos, delivered the way your agent works best.
Choose your context delivery mode
Different tasks need different approaches. Pick the right mode for the job.
| Mode | Description | Token Cost | Latency | Best For |
|---|---|---|---|---|
| Summary | Prompt-optimized key points and essential APIs | Low (~2-4K) | Fast | Quick lookups, broad exploration |
| Full Pack | Complete documentation compiled into single artifact | High (~20-50K) | Medium | Deep implementation, offline use |
| Vector Q&A | Ask questions, get cited answers from docs | Medium (~1-3K) | Fast | Specific questions, troubleshooting |
| Blindspot | Compare model knowledge vs current docs | Low (~1-2K) | Medium | Avoiding hallucinations |
- Summary
- Prompt-optimized key points and essential APIs
- Full Pack
- Complete documentation compiled into single artifact
- Vector Q&A
- Ask questions, get cited answers from docs
- Blindspot
- Compare model knowledge vs current docs
Know what your model doesn't know
LLMs are frozen in time. The blindspot report compares your model's likely training cutoff with the current state of documentation.
What it compares:
- Model's likely knowledge cutoff (declared or inferred)
- Latest docs and package versions
- Changed, deprecated, or new APIs
What it outputs:
- Staleness score (0-100)
- Known-changed APIs list
- Recommended guardrails
- Suggested prompt patch
23 breaking changes detected since the model's training data.
Match exact package versions
Get docs for the version you're actually using, not just "latest".
Lockfile Detection
Automatically detect versions from package-lock.json, yarn.lock, or requirements.txt.
Version History
Access documentation for any historical version, not just the latest release.
Diff Comparison
See what changed between versions to understand migration paths.
Ground with live web content
When documentation isn't enough, fetch live web pages with controlled crawling.
Crawl Controls
- Domain allowlist
- Max depth and page limits
- Rate limiting
- robots.txt compliance
Content Extraction
- Main content selection
- Code blocks preserved
- Citations included
- Provenance trail
Index anything
From npm packages to private repos to local files.
npm Packages
Full npm ecosystem with version-aware indexing.
PyPI Packages
Python packages with docstring extraction.
Public Repos
Any public GitHub repository.
Private Repos
Secure access with OAuth tokens.
Websites
Any crawlable URL or domain.
Local Files
CLI-based local sync with watch mode.
Semantic Search
Vector embeddings for intelligent retrieval.
RAG Pipeline
Retrieval-augmented generation ready.
Full context without flooding context windows
Generate agent-specific context files that load at startup.
Thin
Routing + tool usage only. Minimal token footprint.
~500 tokens
Medium
Key APIs + patterns. Balanced coverage.
~2,000 tokens
Thick
Full pack references + pointers. Maximum knowledge.
~5,000 tokens
Supports: GEMINI.md | CLAUDE.md | AGENTS.md
Resume work across agents
Save conversation histories, plans, and referenced sources. Pick up where you left off.
Conversation Snapshots
Save the full context of a session including tool traces.
Saved in Repo
Context is stored in your repository, automatically shared with teammates.
Cross-Agent
Start in Claude, continue in Gemini. Context transfers seamlessly.
Always current, automatically
Never work with stale docs again. Sources stay synced.
Scheduled Sync
Hourly, daily, or custom schedules for automatic updates.
Webhook Triggers
Instant sync on push events from GitHub or npm publish.
Change Alerts
Get notified via Slack, email, or webhook when docs change.
Build custom agents and workflows
Create reusable workflows, guardrails, and policy checks for your team.
New Repo Setup
Migration Helper
Release Audit
Security Check
Define allowed tools, redaction rules, and approval gates. Available on Pro and Enterprise plans.
Product FAQ
SotaDocs offers Summary (low token context), Full Pack (comprehensive documentation compilation), and Vector Search (precise question-answering with citations) modes.
Skills.md files are project-specific instruction sets (GEMINI.md, CLAUDE.md) that give agents instant knowledge of your tech stack without per-request overhead.
It analyzes your model's likely training cutoff against live documentation to flag changed or deprecated APIs, preventing the agent from using outdated logic.
Yes. While we focus on MCP, we provide a full REST API for custom integrations, allowing you to feed context into any proprietary internal tool.
You can configure sync schedules (hourly, daily, or weekly) or trigger updates via webhooks whenever your internal documentation changes.
Ready to try SotaDocs?
Get Started for Free. No credit card required.