Sourcebot

What do they actually do

Sourcebot is a self-hosted code-understanding platform you deploy inside your own environment. It indexes many repositories across multiple code hosts (GitHub/GitLab/Bitbucket and others) and provides two primary interfaces: a fast cross-repo search/navigation UI and an “Ask Sourcebot” Q&A that answers plain‑language questions about your code with inline citations to the exact files and lines used to form the answer (docs, repo, demo).

The product is designed to keep proprietary code and queries private: it runs on your infrastructure (containerized deployment) and the documentation emphasizes that no data leaves the deployment, with setup guides and telemetry controls provided (site, docs). For automated use cases, Sourcebot also exposes an MCP (Model Context Protocol) server that packages relevant code context for external agents/LLMs so they can operate over large codebases without loading entire repositories into model context (docs, MCP example).

Who are their target customer(s)

• Large enterprise engineering orgs with strict data residency/security needs: They need cross-repo search and code understanding across thousands of repos but cannot use SaaS tools due to privacy, compliance, or scale constraints. A self-hosted deployment where no data leaves their network addresses these constraints (docs, site).
• Internal developer productivity/platform teams: Engineers lose time hunting for definitions, usages, and cross‑repo references via brittle, per-repo tools. They want fast cross‑repo search and NL Q&A with citations that link directly to source lines (demo, docs).
• Teams building AI agents or LLM integrations: Agents hit context limits and pull in irrelevant code. They need a reliable way to deliver only the necessary, cited code snippets to an agent via a context server like MCP (docs, MCP example).
• Security/compliance/legal teams: They require auditability, telemetry controls, and license clarity, and want to minimize data exposure and vendor lock‑in. Self‑hosting, telemetry controls, and the project’s Fair Source license change address these concerns (site, license post).
• Engineering managers/onboarding owners: New hires face long ramp times and scattered tribal knowledge across many repos. Cross-repo navigation and “Ask” answers with inline citations help reduce onboarding overhead (docs, demo).

How would they acquire their first 10, 50, and 100 customers

• First 10: Run white‑glove, in‑network pilots for platform/security teams sourced from YC intros and active demo/GitHub users; time‑box each pilot with a clear success metric and hands‑on setup so they convert into references.
• First 50: Package a standardized 2–4 week private POC (templates for SSO/identity, clear conversion pricing) and hire a sales engineer plus a customer‑success lead to run multiple pilots in parallel and codify fixes.
• First 100: Publish turnkey installers and compliance artifacts (SSO/RBAC, audit logs), launch a small partner/reseller program with SRE and LLM/agent integrators, and add SDR/inbound plus referral incentives to scale without bespoke engineering on every deal.

What is the rough total addressable market

Top-down context:

Depending on source, there are roughly 20.8M professional developers worldwide (JetBrains) to 47M total developers including non‑professionals (SlashData summary). Even if only 10–20% work in environments that require self‑hosted code tools, that implies a multi‑million seat market for private code search/understanding. Separately, 61% of developers report spending >30 minutes per day searching for answers/solutions, underscoring the productivity pain (Stack Overflow 2024).

Bottom-up calculation:

Assume the initial ICP is privacy‑sensitive orgs representing ~10% of professional developers (≈2.1M seats from 20.8M). At an average $300 per seat per year for private code search/understanding and agent context, this implies ≈$630M TAM for the self‑hosted segment.

Assumptions:

• 10% of professional developers work in environments requiring self‑hosted, private code understanding tools.
• Average fully loaded price ≈$300 per seat per year for enterprise self‑hosted deployment.
• One seat per developer in the target segment; excludes adjacent use cases beyond engineering seats.

Who are some of their notable competitors

• Sourcegraph (Code Search + Cody): Mature enterprise code search with self‑hosted options and an AI assistant (Cody) for understanding large codebases; widely used at scale (product page).
• GitHub Enterprise Server (Code Search): GitHub provides advanced code search; on GitHub Enterprise Server, admins manage code search indices for self‑hosted instances—common for orgs fully on GHE (features, GHE search admin docs).
• Glean: Enterprise search/AI assistant that includes code search and agent actions; supports connecting to GitHub/GitLab and provides agent-accessible code context (code search docs, agent/MCP docs).
• OpenGrok: Open‑source, self‑hosted source code search and cross‑reference engine used for large code trees—often a baseline alternative for private deployments (GitHub, overview).
• Codeium (Windsurf): AI coding assistant with enterprise deployment options (including hybrid/self‑hosted components) and multi‑repo context features; overlaps on code understanding and private deployment needs (enterprise, hybrid deployment).