Tessl | Ry Walker Research

Key takeaways

Tessl raised $125M before launching a product — a $25M seed (boldstart, GV) and a $100M Series A led by Index Ventures with Accel — at a reported ~$750M valuation (November 2024)
It is an agent-enablement and spec layer, not an agent fleet — the Tessl Framework and Registry make existing agents (Claude Code, Cursor, Copilot, Gemini) more reliable via specs, skills, and evals
The public registry lists 3,000+ skills with security scanning, version management, install policies, and audit logs; named customers include Cisco and HashiCorp (IBM)
A March 2026 partnership with Snyk attaches a Snyk security score to every public skill in the registry — visible on skill pages, in search, and at CLI install time
Pricing is not publicly listed; the Framework remains in closed beta while the Registry has been in open beta since September 2025

FAQ

What is Tessl?

Tessl is an agent-enablement platform — a spec-driven development framework and a governed registry of agent skills that plug into existing AI coding agents like Claude Code, Cursor, Copilot, and Gemini.

How much does Tessl cost?

Pricing is not publicly listed as of June 2026. The Spec Registry is in open beta and the Framework is in closed beta; enterprise governance features are sold directly.

Does Tessl have its own coding agent?

No — Tessl deliberately sits above the agents. It supplies specs, skills, tests, and governance to whatever agent a team already uses, rather than running an agent fleet itself.

How is Tessl different from Factory AI?

Factory sells autonomous agents (Droids) that do the work; Tessl sells the spec and skill layer that makes any agent's work more reliable and governable.

Executive Summary

Tessl is Snyk founder Guy Podjarny's bet that the bottleneck in AI coding is not the agent but the context around it. The company sells an agent-enablement layer, not an agent fleet: the Tessl Framework captures intent as structured specs with tests as guardrails, and the Tessl Registry distributes versioned, security-scanned "skills" into the agents teams already run — Claude Code, Cursor, Copilot, and Gemini. ^[1] Its tagline is blunt about the thesis: "Skills are the new code. Treat them that way." ^[1]

The funding precedes the product by an unusual margin: $125M across a $25M seed (boldstart, GV) and a $100M Series A led by Index Ventures with Accel, announced November 2024 — before any public launch — at a reported ~$750M valuation. ^[2]^[3] The Registry entered open beta in September 2025, the public skills registry lists 3,000+ skills as of June 2026, and named customers include Cisco and HashiCorp (IBM). ^[4]^[1] In March 2026 Tessl partnered with Snyk to attach a security score to every public skill in the registry — surfaced on skill pages, in search results, and at CLI install time. ^[5] An honest categorization caveat: Tessl belongs in the model-agnostic agentic engineering conversation as the spec/governance layer those platforms plug into, not as a platform that runs agents itself.

Attribute	Value
Company	Tessl
Founder	Guy Podjarny (founder of Snyk, former Akamai CTO) ^[2]
Founded	2024 (came out of stealth November 2024) ^[2]
Funding	$125M total — $25M seed (boldstart, GV) + $100M Series A (Index, Accel) ^[2]
Valuation	~$750M reported (November 2024) ^[3]

Product Overview

Tessl's pitch starts from a diagnosis: "AI coding tools are unreliable" — hallucinated APIs, version blur, and unintended side effects. ^[4] Its answer is spec-driven development (SDD): capture what the software should do in structured spec files that live in the codebase as long-term memory, pair them with tests as hard guardrails, and feed them to whatever agent does the typing. In Podjarny's words, "Spec-driven development gives agents the information they need about both what and how you want them to build, bolstered by tests and hard guardrails." ^[4]

The second product is distribution. The Tessl Registry is effectively a package manager for agent skills and specs — a shared, governed knowledge base "so good work travels instead of duplicating." ^[1] It launched with more than 10,000 pre-built specs covering open-source libraries, and the public skills registry lists 3,000+ searchable skills as of June 2026. ^[4]^[1]

Key Capabilities

Capability	Description
Spec-Driven Framework	Specs with three parts — component description, capabilities linked to tests, API definition; `@generate` creates code from specs, `@describe` documents existing code ^[4]
Skills Registry	3,000+ searchable skills with version management and contribution governance ^[1]
Security & Policy Gating	Every public skill carries a Snyk security score (March 2026 partnership) — shown on skill pages, in search, and at CLI install; admins set install policies and mandate org-standard skills ^[5]^[1]
Usage Specs	Encode team/org rules — tech stack details, internal libraries, APIs, security policies ^[4]
Visibility & Evals	Three-layer visibility (published, project coverage, real activation) plus evaluation-backed skill improvements ^[1]
Agent Integrations	Plugs into Claude Code, Cursor, Copilot, and Gemini ^[1]

Product Surfaces / Editions

Surface	Description	Availability
Tessl Registry	Public + private registry of skills and specs	Open beta (since September 2025) ^[4]
Tessl Framework	Spec-driven development workflow and tooling	Closed beta ^[4]
Enterprise governance	Install policies, ownership/access controls, audit logs, inventory	Sold directly ^[1]

Technical Architecture

Tessl is deliberately agent-agnostic and model-agnostic — it does not run the agent or pick the model. Specs and skills are files and packages that existing agents consume, which means Tessl inherits whatever models its supported agents use. ^[1]

Key Technical Details

Aspect	Detail
Deployment	Cloud registry + in-repo spec files; specs live "in the codebase as long-term memory" ^[4]
Model(s)	None of its own — works through Claude Code, Cursor, Copilot, Gemini ^[1]
Integrations	The four major coding agents above; org-internal libraries/APIs via usage specs ^[4]
Open Source	Platform is proprietary; company positions its ecosystem as "open and composable" ^[2]

Strengths

Founder pedigree — Guy Podjarny built Snyk into a developer-security standard; Tessl applies the same registry-plus-governance playbook to agent skills ^[2]
War chest before launch — $125M from Index, Accel, GV, and boldstart provides multi-year runway for a pre-revenue-stage company ^[2]
Agent-agnostic by design — value survives agent churn; switching from Cursor to Claude Code doesn't strand your specs and skills ^[1]
Governance depth — security scanning, policy gating, audit logs, and activation visibility target exactly what enterprises complain is missing from ad-hoc skill/prompt sharing ^[1]
Snyk-backed skill security — since March 2026 every public registry skill carries a Snyk security score; the underlying research found 36% of skills on a rival marketplace (ClawHub) contained prompt-injection techniques, which is the threat this gating addresses ^[5]
Early enterprise logos — Cisco and HashiCorp (IBM) practitioners are quoted on the homepage ^[1]
Registry scale — launched with 10,000+ pre-built library specs; 3,000+ skills publicly searchable as of June 2026 ^[4]^[1]

Cautions

Not an agent platform — Tessl does no work itself; teams still need to buy or adopt the agents, and agent vendors (Anthropic, GitHub, AWS) are building competing native skill/spec ecosystems
Spec-driven development is contested — the loudest HN thread on SDD is largely skeptical, with commenters calling markdown specs unvalidatable and the workflow "waterfall, but with AI" ^[6]
Beta maturity — the Framework is still closed beta and the Registry open beta roughly nine months after launch; this is early-stage software with a late-stage valuation ^[4]^[3]
No public pricing — as of June 2026 there is no published pricing page, making evaluation and budgeting opaque ^[1]
Valuation-to-proof gap — ~$750M valuation was set before product launch; adoption metrics (beyond skill counts and two named customers) are not publicly disclosed ^[3]
Thin direct community signal — Tessl's $125M raise drew only 4 comments on Hacker News; most community discussion is about the SDD category, not the product ^[7]

What Developers Say

As of June 2026, substantive community discussion of Tessl the product remains thin — the funding announcement drew 4 HN comments ^[7] — and the freshest practitioner signal comes from Tessl's own AI Native DevCon (London, June 1–2, 2026), where Podjarny's keynote argued for treating skills as versioned, tested, owned software. ^[8] The spec-driven development category it champions drew a lively, largely skeptical 128-point HN thread (which covers Kiro, Spec-Kit, and Tessl). ^[6]

"Really, we are doing waterfall, but with AI, now?" — constantcrying, Hacker News ^[6]

"This focus on markdown specs is the dumbest thing. Have a spec DSL that can be validated and transformed into real code." — CuriouslyC, Hacker News ^[6]

"The code is the spec—the only spec that matters. The only way of describing behavior everyone can agree on the meaning of." — conartist6, Hacker News ^[6]

"We've been going all-in on spec-as-source... it's really the most interesting use of specs, but it's also challenging to get off the ground." — fabianlindfors, Hacker News ^[6]

"I keep a spec of the solution developed, and include it in every request sent to the ai, and it yields good results in my case." — raphinou, Hacker News ^[6]

Pricing & Licensing

Tier	Price	Includes
Registry (open beta)	Not publicly listed	Public skill search, install into supported agents ^[1]
Framework (closed beta)	Not publicly listed	Spec-driven development workflow ^[4]
Enterprise	Custom (direct sales)	Install policies, security scanning, access controls, audit logs ^[1]

Licensing model: Proprietary platform; no published pricing as of June 2026 — expect direct/enterprise sales motion.

Hidden costs: Tessl is additive — teams still pay for the underlying agents (Claude Code, Cursor, Copilot, Gemini) and their token usage; spec authoring and maintenance is its own ongoing labor.

Competitive Positioning

Categorization caveat: Tessl is an agent-enablement/spec layer, not an agent fleet. It competes with other spec/skill ecosystems for mindshare, and complements (rather than replaces) the platforms that actually run agents.

Direct Competitors

Competitor	Differentiation
GitHub Spec Kit	Free, open-source SDD toolkit; Tessl adds a governed registry, security scanning, and enterprise controls
AWS Kiro	Spec-driven IDE — bundles its own agent; Tessl is agent-agnostic across four major agents
Anthropic Skills ecosystem	Native skills for Claude Code only; Tessl is cross-agent with org governance
Factory AI	Factory sells autonomous Droids that do the work; Tessl sells the spec/skill layer any agent consumes
Tembo	Tembo orchestrates agent fleets against tickets; Tessl governs the knowledge those agents run on

When to Choose Tessl Over Alternatives

Choose Tessl when: You run multiple coding agents and need one governed, security-scanned source of truth for skills, specs, and org standards
Choose Spec Kit/Kiro when: You want SDD for free (Spec Kit) or bundled with an IDE agent (Kiro) without an enterprise registry
Choose Factory AI when: You want to buy agents that execute work, not a layer that informs them
Choose Tembo when: You need fleet orchestration — routing tickets to agents and reviewing output — rather than skill governance

Ideal Customer Profile

Best fit:

Enterprises running several coding agents (Claude Code + Copilot + Cursor) that need consistent standards across all of them
Platform engineering teams who want to mandate org-standard skills with audit trails and security scanning
Organizations with internal libraries/APIs that agents repeatedly get wrong — usage specs encode those rules once
Security-conscious shops that liked the Snyk model of developer tooling with governance attached

Poor fit:

Teams that want an agent that does the work — Tessl supplies context, not execution
Solo developers and small teams; a CLAUDE.md/AGENTS.md file plus free Spec Kit covers most of the need
Buyers who require public pricing and GA software today
Organizations standardized on a single agent vendor's native skill ecosystem

Viability Assessment

Factor	Assessment
Financial Health	Strong — $125M raised (Index, Accel, GV, boldstart) at a reported ~$750M valuation ^[2]^[3]
Market Position	Early — credible founder and logos (Cisco, HashiCorp/IBM), but the spec/skill-layer category is unproven and contested ^[1]
Innovation Pace	Active — registry launch Sept 2025, Snyk security-score integration March 2026, skills/evals/governance shipping through 2026 ^[4]^[5]
Community/Ecosystem	Thin but forming — 3,000+ registry skills and a company-run conference (AI Native DevCon London, June 1–2, 2026), yet little independent product discussion ^[1]^[8]^[7]
Long-term Outlook	Uncertain-positive — runway is long; the open question is whether the governed-skills layer becomes a standalone market or a feature of agent platforms

Tessl has the rare combination of a repeat founder who built the analogous company (Snyk: registry-adjacent developer security with enterprise governance) and enough capital to outlast the category's definition phase. The risk is structural, not financial: agent vendors are racing to own skills natively, and SDD skepticism in the developer community is real.

Bottom Line

Tessl is the most heavily capitalized bet that the durable layer in AI coding is specs, skills, and governance — not the agents themselves. If you accept that framing, its registry-with-governance design (Snyk-scored security scanning, policy gating, audit logs, cross-agent reach) is the most enterprise-complete version of the idea shipping today. But buyers should be clear-eyed: this is beta software, with no public pricing, in a category that loud parts of the developer community consider re-warmed waterfall.

Recommended for: Multi-agent enterprises that need governed, security-scanned, org-wide standards for what their coding agents know and do.

Not recommended for: Teams shopping for an agent or agent fleet — Tessl doesn't execute work — or small teams for whom free options (Spec Kit, plain AGENTS.md files) suffice.

Outlook: Watch three things through 2026: whether the Framework reaches GA with public pricing, whether registry skill counts translate into disclosed customer adoption beyond Cisco and HashiCorp, and whether agent vendors' native skill ecosystems commoditize the cross-agent registry before it becomes the standard.

Research by Ry Walker Research • methodology

Sources