Terfyn

A statically analyzable, capability-oriented execution platform for nondeterministic programs. Review the authority granted to agents as a plan diff — before they run.

Terfyn bounds and diffs that grant. It does not verify what remote systems do with it.

Architecture

Source graph → validate / plan → SQLite desired state → apply → engine → tools + models → trace / logs / audit.

flowchart LR
  Source[Source graph] --> VP["validate / plan"]
  VP --> SQLite[(SQLite desired state)]
  SQLite --> Apply[apply]
  Apply --> Engine[engine]
  Engine --> Tools[tools + models]
  Engine --> Trace[trace / logs / audit]

Expanded diagram, plan-time bounds, and closed-world caveats: docs/architecture.md. Product spec: docs/DESIGN_DOC.md.

The differentiator: plan-time bounds on authority

This is not another orchestrator (Temporal, Dagger, LangGraph). Those schedule work. Terfyn's direction is a plan-time effect bound — a sound static upper bound on what an autonomous agent can do, reviewable as a diff (#189 / #191). terfyn plan prints that bound and the authority delta vs stored deployment state.

Today terfyn plan diffs permissions, approvals, models, budgets, C1 risk items, and effect/capability/authority against SQLite desired state:

Plan: 0 to add, 3 to change, 0 to delete
~ update Agent/reviewer
    spec.model: "mock/gpt-4" -> "mock/gpt-4o"
    spec.tools.1:  -> "github"
~ update Policy/default
    spec.approvals.requiredFor.0: "tool.helper.echo" -> "tool.github.issues.write"
    spec.execution.maxTotalCostUsd: 3 -> 10
~ update Tool/github
    spec.permissions.allow.1:  -> "issues.write"

Risk delta:
high:
- [high] approval_removal: Approval requirements removed for "tool.helper.echo" (Policy/default).
- [high] budget_relaxation: Cost ceiling increased (Policy/default).
- [high] permission_widening: New write-like tool permission "issues.write" added (Tool/github).
- [high] tool_surface_change: Agent tools list gained write-like tool "github" (Agent/reviewer).
medium:
- [medium] model_change: Agent model changed (Agent/reviewer).

When the graph declares tool operations, plan also prints the effect bound and an authority delta (bound(desired) vs bound(deployed)). Capability changes and effect changes are separate lines; AUTONOMOUS WIDENED means a nondeterministic agent's action space grew — including when a new grant does not add named effects.

Effect bound (Workflow/pr-review):
high:
- [high] effect_bound: github.write       autonomous  Agent/reviewer may select tool.github.post_comment
medium:
- [medium] effect_bound: github.read        static      step fetch_pr

Capability delta:
Agent/reviewer
+ tool.github.post_comment

Authority:
  static      -> unchanged
  autonomous  -> WIDENED

Agents and workflows are authored in .agent, the surface syntax fixed by ADR 002; the loader compiles .agent (type/effect checking + argument rebind) into the resource graph. Workflows run end-to-end, including conditionals, loops, and dynamic fan-out (#199/#259): a control-flow workflow lowers to the execution IR, is pinned into the deployment snapshot, and runs on the execir interpreter (see examples/agent-control-flow). YAML is the compilation output and interchange format (ADR 003): the loader still accepts it (so machine-generated resources and the existing fixtures work), and terfyn export --format yaml materializes the compiled graph on demand. Lead on capability, not format.

Flagship: incident triage

Start here: examples/incident-triage — an offline agent that can page, read logs, and file a ticket, but cannot restart a service unless policy approvals.requiredFor includes tool.restart.restart and that uses string is pre-approved with --approve tool.restart.restart. Inner agent-loop tools do not HITL; unapproved terfyn run fail-closes with exit 5 (approval_required). With --approve it completes and audit verify passes.

Other walkthroughs: policy-blocked PR review in examples/pr-review-demo (no API keys); live GitHub read/write with a mock reviewer in examples/pr-review-github; the same flow with OpenAI gpt-4o-mini plus GitHub Actions in examples/pr-review-github-actions (PR workflow; optional manual owner/repo/number).

Why this exists

Most agent stacks bury prompts, tool wiring, and permissions in application code. That makes it hard to answer: Is this config valid? What authority changed? What are we about to grant? What actually ran? Did policy allow it?

Terfyn is a small Go CLI (terfyn) plus a resource graph so teams can:

Review capability diffs (plan) before changes land
Track deployment state separately from runtime traces
Enforce policies (budgets, approvals, tool rules) at execution time
Stay local-first while the architecture leaves room for a future remote control plane

Mental model

Idea	Analogy
Desired resources in Git	GitOps
`plan` / `apply` / drift	Terraform
Typed resources (`Project`, `Workflow`, `Policy`, …)	Kubernetes-style API
Tool and IO contracts	OpenAPI-style explicitness

Compared with adjacent tools

Terfyn is the declarative governance/config layer for agent systems — not a replacement for durable-execution engines or code-first agent runtimes.

Capability	This project	OpenAI Agents SDK	LangGraph	Temporal	Terraform
Role	Governance/config: versioned resources, plan/apply, policy	Code-first agent runtime	Code-first graph orchestration	Durable workflow execution	Infrastructure as code
Durable execution / distributed scheduling	No	No	Optional checkpointers; not a durable-execution engine	Yes	N/A
Code-first agent runtime	No (resource graph; `.agent` authoring, YAML compilation output)	Yes	Yes	Workflow SDK, not an agent runtime	No
Desired-state plan / apply	Yes (`terfyn plan` / `apply` vs SQLite)	No	No	No	Yes
Plan-time effect bound	Shipped (#189 / #190 / #191): bound over the callable operation set, including autonomous tool selection; `terfyn plan` prints the bound and authority delta. No listed comparable.	No	No	No	No

The bound is not over what those operations do at the far end; the trust anchor is human review of the tool manifest. Manifest pin (#204) is not enforced today (MCP tools/list can still expand the world). terfyn plan diffs permissions, approvals, models, budgets, C1 risk items, and effect/capability/authority.

Features (MVP today)

terfyn init — scaffold a .agent-led project (main.agent workflow plus project.yaml, policies, tools)
terfyn export --format yaml — materialize the compiled resource graph as YAML on demand (nothing written to disk by default; --output DIR writes a loadable project)
terfyn fmt — format .agent sources to canonical form (and normalize project YAML)
terfyn validate — load project, apply project defaults (spec.defaults), then environment overlays (-e / --env, Environment resources §7.6), then validate graph, schemas, and references; runs policy lint (ungated sensitive tools, invalid HITL config, etc.) as advisory output — use --strict to exit 2 on high-severity lint findings (fail-closed safety metadata still gates at run even when lint passes)
terfyn plan — diff desired graph vs SQLite deployment state; risk hints including policy lint, effect bound, and authority delta; JSON/YAML output includes policyLint, deploymentBaseline, effectBound, and authority
terfyn apply — persist plan (TTY confirm or --auto-approve / TERFYN_AUTO_APPROVE); optimistic concurrency — if the deployment store changed after the plan snapshot (e.g. another process applied the same --state file while this run waited at the prompt), apply fails with exit code 3; re-run plan then apply
terfyn run — execute a workflow locally; JSON Schema for inputs where configured; policy gates pause for human-in-the-loop (HITL) approval when a tool call requires it
terfyn logs — read trace events from SQLite (--run, --workflow, or recent runs)
terfyn audit verify — re-walk hash-linked trace chains and detect tampering (see docs/AUDIT_CHAIN.md)
Tools — native, http, mock, and mcp — MCP supports stdio (subprocess) or streamable HTTP (spec.mcp.transport: http, url, optional headers with env: tokens)
Project defaults — besides model and policy, optional runtime flows to spec.runtime on agents/workflows when omitted (MVP: local or unset; see spec validation)
Output — table, JSON, or YAML (-o / --output)
State — single SQLite file (default .agentic/state.db under the project root; override with --state)
Tests — unit/integration coverage, golden CLI output tests, end-to-end init → … → logs in test/integration

See section 18 (MVP) and section 19 (End Goal) in docs/DESIGN_DOC.md for the full included/excluded list.

Quick start

Prerequisites

From source: Go 1.22+
From a release binary: no Go toolchain; put terfyn (or terfyn.exe) on your PATH after extracting the archive

Build

git clone https://github.com/LAA-Software-Engineering/terfyn.git
cd terfyn
make build   # writes bin/terfyn

Or: make install / go install ./cmd/terfyn (honours GOBIN / GOPATH/bin; ensure that directory is on PATH).

Prebuilt binaries

GitHub Releases ship terfyn for common platforms (.tar.gz on Linux/macOS, .zip on Windows) plus SHA256SUMS.txt. Pick the archive that matches your machine, for example:

Platform	Asset suffix
Linux x86_64	`linux-amd64.tar.gz`
Linux arm64	`linux-arm64.tar.gz`
macOS Intel	`darwin-amd64.tar.gz`
macOS Apple Silicon	`darwin-arm64.tar.gz`
Windows x86_64	`windows-amd64.zip` (contains `terfyn.exe`)

terfyn version reports the release tag (e.g. v0.1.4).

Releases are created automatically when changes land on main, using a patch semver bump over the latest vMAJOR.MINOR.PATCH tag (merges that only touch Markdown or the root Makefile do not trigger a release). To cut minor or major bumps on demand, run the Release workflow manually (Actions → Release → Run workflow) and choose the bump type.

Create a project and run the loop

From the repo root (or anywhere):

terfyn init my-agent-system
terfyn validate --project my-agent-system
terfyn plan   --project my-agent-system
terfyn apply  --project my-agent-system --auto-approve
terfyn run    workflow/hello --project my-agent-system
terfyn logs   --project my-agent-system --workflow hello
terfyn audit verify --project my-agent-system --run <run-id>
terfyn inspect --web --project my-agent-system   # read-only local UI on http://127.0.0.1:8787

inspect --web binds to localhost only and opens the state DB read-only. Avoid running it while terfyn run is writing the same SQLite file (you may see database is locked without WAL); use it when runs are idle or on a copy of the DB.

Authoring: `.agent` plus `project.yaml`

Agents and workflows are authored in .agent (grammar reference); .agent files anywhere under the project root are discovered and compiled automatically. Tools, policies, and project configuration stay YAML (the interchange format). After terfyn init my-agent-system, my-agent-system/main.agent is:

workflow hello() {
    helper.echo(message: "hello")
}

and my-agent-system/project.yaml holds the config — a Project resource with spec.imports listing the YAML resources (policies, tools); .agent sources are not imported, they are discovered:

apiVersion: agentic.dev/v0
kind: Project
metadata:
  name: my-agent-system
spec:
  imports:
    - ./policies/default.yaml
    - ./tools/helper.yaml
  defaults:
    policy: default
    model: openai/gpt-4o-mini
    runtime: local
  providers:
    models:
      openai:
        type: openai
        apiKeyFrom: env:OPENAI_API_KEY
      # Optional: Claude via Messages API (set ANTHROPIC_API_KEY and use e.g. defaults.model: anthropic/claude-sonnet-4-20250514)
      # anthropic:
      #   type: anthropic
      #   apiKeyFrom: env:ANTHROPIC_API_KEY

To see the compiled resource graph as YAML — for inspection or handoff to another tool — run terfyn export --format yaml (it prints to stdout; nothing is written to disk unless you pass --output DIR). YAML remains valid ingress, so you can still author or generate resources directly in it.

Field-by-field rules, extra kinds, env overlays, MCP HTTP tools, and defaults.runtime are in docs/DESIGN_DOC.md. See docs/EXAMPLES.md for Anthropic fragments, MCP over HTTP, and structured-output notes.

Notes:

init creates my-agent-system/ with apiVersion: agentic.dev/v0 resources and a hello workflow (native echo tool only — no network).
apply in non-interactive environments needs --auto-approve or TERFYN_AUTO_APPROVE=1.
run HITL: gated tool calls exit with Status: interrupted (exit 0). Resume with --resume <run-id> --decision approve|reject|edit|switch (use --decision-edit-json / --decision-switch-target when needed), or skip prompts with --auto-approve / TERFYN_AUTO_APPROVE=1. Pre-approve a specific call with repeated --approve <uses>. Set TERFYN_HITL_ACTOR to attribute decisions in trace logs.
Policy.spec.hitl.interruptOn keys are Tool metadata.name values; they configure review options (edit rules, switch targets) for calls already gated by approvals.requiredFor or safety metadata — they do not gate tools on their own.
run stores traces in the same SQLite file used for plan/apply (default .agentic/state.db under --project). Optional OTLP export (spec.telemetry, off by default) is additive only — see docs/OTEL.md. When enabled you need serviceName plus either consoleExport: true or an endpoint (https://… or env:VAR, e.g. env:OTEL_EXPORTER_OTLP_ENDPOINT). Export that variable before run if you use env:; if it is missing or the collector is unreachable, terfyn logs a warning, skips OTLP, and the workflow still completes (SQLite traces unchanged).
If spec.traces.retentionDays is a positive integer, runs older than that many UTC calendar days (by runs.started_at) are deleted lazily on run and logs (child trace rows cascade). Unset or non-positive means no pruning.
Trace payload redaction (issue #110): before SQLite storage, event JSON is sanitized, key-redacted, and size-capped. Defaults mask common secret key names (substring match on map keys). Optional project knobs:
- spec.traces.redactKeys / maxPayloadBytes — merged with defaults; also available under spec.traces.redaction together with maxDepth, maxStringChars, and maxBytes (max bytes for binary previews in sanitized values, not the overall JSON cap).
- Stored events may show [REDACTED], payload_truncated / preview, or depth/binary placeholders in logs / inspect --web.
Use logs --run <id> after a run if you want a single run’s trace (IDs are printed by run).

Global flags (common)

Flag	Purpose
`--project <path>`	Project root (default `.`)
`--state <path>`	SQLite file override
`-e` / `--env`	Environment overlay name
`-o` / `--output`	`table`, `json`, or `yaml`
`--no-color`	ASCII-friendly validate output

Exit codes are summarized in section 11.2 of docs/DESIGN_DOC.md (0 success, 2 validation, 3 plan/apply conflict when deployment state changed after plan or resolved config drifted before run, 4 execution, 5 policy denial, …).

User-local config (per-developer overrides)

Config is resolved in this order (highest wins): CLI flags → environment overlay (-e) → project YAML → user-local → built-in defaults.

Optional user-local files (git-ignored, strict YAML — typos fail validate):

Path	Scope
`$XDG_CONFIG_HOME/terfyn/config.yaml` or `~/.config/terfyn/config.yaml`	Global per-user defaults (`defaults`, `state`, `providers`, `traces`, `telemetry`)
`.agentic/local.yaml` under `--project`	Project-scoped overrides (same fields; wins over the global file)

validate, plan, and apply write .agentic/resolved-config.json (digest of the resolved graph + env + state path). run rejects drift from that snapshot with exit 3 — re-run validate or plan after changing config.

Repository layout (high level)

Path	Role
`cmd/terfyn`	CLI entrypoint
`internal/cli`	Cobra commands, flags, golden tests
`internal/spec`	YAML types, normalize, validate
`internal/config`	Layered config resolution, immutable snapshot
`internal/project`	Load project + imports
`internal/plan`	Planner and risk summary
`internal/apply`	Apply plan to deployment store
`internal/engine`	Workflow execution
`internal/policy`	Policy evaluation
`internal/state/sqlite`	SQLite deployment + runtime/trace tables
`internal/audit`	Tamper-evident hash chain for trace events (issue #116)
`test/integration`	End-to-end CLI flow tests
`docs/DESIGN_DOC.md`	Spec, CLI UX, architecture, roadmap
`docs/GITHUB_ACTIONS.md`	Running `terfyn` from GitHub Actions (tokens, exit code 5, template path)
`examples/pr-review-github-actions/`	Full `gpt-4o-mini` project; PR workflow `.github/workflows/terfyn-pr-review.yml`; optional publish `.github/workflows/terfyn-pr-review-publish.yml`

Development

make defaults to help, which lists targets; the table below mirrors the Makefile (## comments and recipes).

Target	What it does
`help`	Show usage and target list (default goal)
`all`	`fmt` → `vet` → `test` → `build` (handy before a push)
`build`	`go build` → `bin/terfyn`
`install`	`go install ./cmd/terfyn` (`-trimpath`; uses `GOBIN` / `GOPATH/bin`)
`clean`	Remove `bin/` and `coverage.out`
`fmt`	`go fmt ./...`
`verify-fmt`	Fail if `gofmt -l` would list files (matches CI-style formatting check)
`vet`	`go vet ./...`
`test`	`go test ./... -race`
`test-coverage`	Tests with `-coverprofile=coverage.out` and a one-line `go tool cover -func` summary
`check`	`vet` + `test` only (no formatting writes)
`ci`	`verify-fmt` + `vet` + `test` (no build)

CI (.github/workflows/ci.yml) runs Linux, macOS, and Windows on Go 1.22.x, plus Go 1.23.x on Linux, with race and shuffle enabled (workflow steps are defined in YAML, not via make ci).

Updating CLI golden files

When table output is intentionally changed:

GO_UPDATE_GOLDEN=1 go test ./internal/cli/... -run TestGolden_

Roadmap

Near term (MVP hardening)

Recent landings already cover much of “hardening”: plan/apply optimistic concurrency (exit 3 when deployment state drifts), MCP over streamable HTTP as well as stdio, trace retention (spec.traces.retentionDays), defaults.runtime / spec.runtime (MVP local), and clearer defaults vs environment overlay documentation. What is still open for near-term polish:

More diff / drift UX where the design doc calls for it (beyond today’s resource-level diff)
Richer logs filtering (see sections 10.2 and 17.3 in docs/DESIGN_DOC.md); inspect --web covers read-only run/state browsing (#109)
terfyn test-style workflow fixtures (stretch per design doc)

Post-MVP (from design doc section 19)

Modules/registry, remote shared state, reconciliation controllers
Parallel steps, subworkflows, schedules/events
Stronger drift semantics and multi-runtime targets
Deeper approval workflows and multi-tenant controls

The recommended implementation phases are outlined in section 20 of docs/DESIGN_DOC.md.

Documentation

docs/DESIGN_DOC.md — design document v0 (problem statement, spec, CLI, engine, state model, testing strategy, MVP vs end state, section 23 recommendation).
docs/AGENT_LOOP.md — bounded agent tool-calling loop: grants, advertised uses, maxIterations, policy on every inner call, traces, HITL vs exit 5 (issues #160 / #161 / #175).
docs/AUDIT_CHAIN.md — hash-linked trace audit chain and terfyn audit verify (issue #116).
docs/ATTRIBUTION.md — tenant, thread, and actor fields on runs and traces (issue #111).
docs/OTEL.md — optional OTLP trace export alongside SQLite (issue #108).
docs/TESTING.md — terfyn test fixture format; CI gate walkthrough in examples/regression-test.
examples/pr-review-demo/README.md — end-to-end demo: structured review output, traceable run, approval-gated write (validate → plan → apply → run → logs).
examples/regression-test/README.md — terfyn test is green on a gated publish and red after dropping requiredFor (issue #176).
docs/EXAMPLES.md — copy-paste YAML and CLI examples (init, mock vs OpenAI, workflows, environment overlays).
CODE_OF_CONDUCT.md — Contributor Covenant 2.1; participation expectations and reporting.
License: MIT

Contributing

Issues and pull requests are welcome. See CONTRIBUTING.md for local setup, tests, golden updates, and pull request expectations.

Name		Name	Last commit message	Last commit date
.cursor/skills		.cursor/skills
.github		.github
cmd/terfyn		cmd/terfyn
docs		docs
examples		examples
internal		internal
migrations/sqlite		migrations/sqlite
scripts		scripts
test/integration		test/integration
.gitattributes		.gitattributes
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
CODEOWNERS		CODEOWNERS
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
Makefile		Makefile
README.md		README.md
go.mod		go.mod
go.sum		go.sum

Navigation Menu

Uh oh!

Repository files navigation

Terfyn

Architecture

The differentiator: plan-time bounds on authority

Flagship: incident triage

Why this exists

Mental model

Compared with adjacent tools

Features (MVP today)

Quick start

Prerequisites

Build

Prebuilt binaries

Create a project and run the loop

Authoring: `.agent` plus `project.yaml`

Global flags (common)

User-local config (per-developer overrides)

Repository layout (high level)

Development

Updating CLI golden files

Roadmap

Near term (MVP hardening)

Post-MVP (from design doc section 19)

Documentation

Contributing

About

Releases

Used by

Contributors

Languages

Uh oh!

Latest commit

History

Folders and files

Repository files navigation

Terfyn

Architecture

The differentiator: plan-time bounds on authority

Flagship: incident triage

Why this exists

Mental model

Compared with adjacent tools

Features (MVP today)

Quick start

Prerequisites

Build

Prebuilt binaries

Create a project and run the loop

Authoring: .agent plus project.yaml

Global flags (common)

User-local config (per-developer overrides)

Repository layout (high level)

Development

Updating CLI golden files

Roadmap

Near term (MVP hardening)

Post-MVP (from design doc section 19)

Documentation

Contributing

About

Topics

Resources

Code of conduct

Contributing

Stars

Watchers

Forks

Releases

Used by

Contributors

Languages

Authoring: `.agent` plus `project.yaml`